University of Bath

PHD

Investigations into hybrids of carbon nanotubes and organo-metallic molecularsystems

Lewis, Peter

Award date:2014

Awarding institution:University of Bath

Link to publication

General rightsCopyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright ownersand it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.

• Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain • You may freely distribute the URL identifying the publication in the public portal ?

Take down policyIf you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediatelyand investigate your claim.

Download date: 10. Jun. 2020

Investigations into hybrids of carbonnanotubes and organo-metallic

molecular systemsSubmitted by

Peter Rex Lewisfor the degree of Doctor of Philosophy

of the

University of BathDepartment of Physics

November 2013

Copyright

Attention is drawn to the fact that copyright of this thesis rests with the

author. A copy of this thesis has been supplied on condition that anyone

who consults it is understood to recognise that its copyright rests with the

author and that they must not copy it or use material from it except as

permitted by law or with the consent of the author.

This thesis may be made available for consultation within the University Li-

brary and may be photocopied or lent to other libraries for the purposes of

consultation.

Signature of author. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Peter Rex Lewis

Abstract

Endohedral functionalization via supercritical CO2 was undertaken in or-

der to produce encapsulation of organometallic systems that are difficult to

encapsulate otherwise due to either their large size or extreme air sensitivity.

Organometallic molecular systems from the prophyrin and phthalocyanine

families (such as NiPc, ClAlPc and NiTPP) were successfully encapsulated

inside of nanotubes with relatively large diameters (centred around 2 nm).

This was assessed by a combination of high resolution transmission electron

microscopy (HRTEM) and Raman spectroscopy. HRTEM revealed previ-

ously unreported ordering of NiTPP, a large planar molecule, in row-like as-

semblies inside nanotubes of diameters that match best the geometrical size

of the molecule (2 nm), highlighting the role of confinement in promoting as-

sembly. Using both endohedral and exohedral functionalization with NiTPP,

ClAlPc and NiPc molecules provided a set of systems differing by only one

specific parameter (e.g. central ion or body type, or size of the HOMO-

LUMO gap), or comparatively affected the ability to bind to the nanotubes

- the associated changes in the electronic properties of the nanotubes were

revealed by resonant Raman spectroscopy. These changes were interpreted

in terms of ability of the guest molecular species to produce charge transfer

to/from the nanotube, and/or induce structural strain.

Acknowledgements

I would like to thank the following people for the contributions they have

made to my PhD studies: My supervisor Dr. Adelina Ilie for all her support,

mentoring and guidance.

All the academic and support staff of the University of Bath who assisted

me during my studies, in particular, Dr. John Mitchels for his help with

operating the TEM and Raman spectrometer of the Microscopy and Analysis

suite; Dr. Daniel Wolverson of the Physics department for his help and

guidance with my Raman spectroscopy experiments; Dr. Simon Brayshaw

of the Chemistry department for his assistance with my molecular filling

experiments; Dr. Michael Grogan for his help with learning how to use the

software of the supercritical CO2 rig and Wendy Lambson for all her help

and advice in the ordering and use of the chemicals which I have needed

during my studies.

The staff at Exeter University Physics department, in particular, Dr.

Annette Plaut for kindly arranging for me to visit Exeter to carry out some

of my Raman spectroscopy investigations and Ellen Green for setting up the

Raman spectrometer and teaching me how to use it. My family and friends

for supporting me throughout my PhD.

Contents

1 Introduction 1

1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.2 Summary of thesis . . . . . . . . . . . . . . . . . . . . . . . . 4

2 Background and theory 5

2.1 Carbon . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

2.2 Graphene . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

2.2.1 The graphene lattice . . . . . . . . . . . . . . . . . . . 8

2.2.2 The reciprocal lattice of graphene . . . . . . . . . . . . 9

2.2.3 The tight-binding approximation for the energy band

structure . . . . . . . . . . . . . . . . . . . . . . . . . . 11

2.2.4 The electronic band structure of graphene . . . . . . . 15

2.3 Carbon nanotubes . . . . . . . . . . . . . . . . . . . . . . . . 17

2.3.1 Structure . . . . . . . . . . . . . . . . . . . . . . . . . 17

2.3.2 The reciprocal lattice of SWNTs . . . . . . . . . . . . . 21

2.3.3 Electronic structure of SWNTs . . . . . . . . . . . . . 23

2.3.4 Electronic density of states . . . . . . . . . . . . . . . . 25

4

3 The encapsulation of molecular systems by carbon nanotubes

using a supercritical carbon dioxide-based method 28

3.1 The effect of curvature upon the

reactivity and binding capability of SWNTs . . . . . . . . . . 29

3.2 Effects of curvature upon molecular

adhesion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

3.3 Aromatic interaction between molecules and the exterior of

CNTs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

3.4 Effects of curvature upon molecule

diffusion on and in SWNTs . . . . . . . . . . . . . . . . . . . 40

3.5 Molecular encapsulation by SWNTs . . . . . . . . . . . . . . . 41

3.5.1 Encapsulation methods . . . . . . . . . . . . . . . . . . 41

3.6 Molecular encapsulation by SWNTs using a ScCO2 medium . 45

3.6.1 Supercritical fluids . . . . . . . . . . . . . . . . . . . . 45

3.7 ScCO2 induced encapsulation . . . . . . . . . . . . . . . . . . 48

4 Production of hybrids of nanotubes and organo-metallic molec-

ular systems 52

4.1 Synthesis of carbon nanotubes . . . . . . . . . . . . . . . . . . 52

4.1.1 Laser vaporization . . . . . . . . . . . . . . . . . . . . 53

4.1.2 Arc-discharge . . . . . . . . . . . . . . . . . . . . . . . 53

4.1.3 Carbon vapor deposition . . . . . . . . . . . . . . . . . 55

4.2 Purification of carbon nanotubes . . . . . . . . . . . . . . . . 56

4.2.1 Impurities . . . . . . . . . . . . . . . . . . . . . . . . . 56

4.2.2 Thermal oxidation . . . . . . . . . . . . . . . . . . . . 56

4.2.3 Hydrogen peroxide-based oxidation . . . . . . . . . . . 57

5

4.2.4 Acid reflux . . . . . . . . . . . . . . . . . . . . . . . . 57

4.2.5 Purification procedures adopted . . . . . . . . . . . . . 57

4.3 Characterisation of purified nanotubes . . . . . . . . . . . . . 59

4.3.1 Characterisation of Arc SWNTs . . . . . . . . . . . . . 59

4.3.2 Characterisation of CVD SWNTs . . . . . . . . . . . . 60

4.4 Nanotube end-opening . . . . . . . . . . . . . . . . . . . . . . 62

4.5 Supercritical fluid molecular filling

experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

4.5.1 Equipment set-up . . . . . . . . . . . . . . . . . . . . . 63

4.5.2 Configuration A . . . . . . . . . . . . . . . . . . . . . . 64

4.5.3 Configuration B . . . . . . . . . . . . . . . . . . . . . . 65

4.6 Sample production . . . . . . . . . . . . . . . . . . . . . . . . 67

4.6.1 (a) ScCO2 filling of nanotubes from powder . . . . . . 67

4.6.2 Summary of samples produced . . . . . . . . . . . . . . 69

4.6.3 (b) ScCO2 filling of nanotubes from a molecular solution 69

4.6.4 (c) Exohedral functionalisation of SWNTs . . . . . . . 71

4.7 Removal of extraneous molecular

material . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72

5 HRTEM investigations of the internal structure of hybrids

of nanotubes and organo-metallic molecular systems 73

5.1 High Resolution Transmission Electron

Microscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74

5.2 Equipment and experimental methods . . . . . . . . . . . . . 77

5.3 HRTEM of related systems . . . . . . . . . . . . . . . . . . . . 79

6

5.4 Structural characterisation of hybrids of SWNTs and endohe-

dral NiTPP . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81

6 Resonant Raman spectroscopy of filled carbon nanotubes 86

6.1 Introduction to the Raman effect . . . . . . . . . . . . . . . . 86

6.1.1 Raman-active molecules - a classical treatment . . . . . 86

6.1.2 Photonic scattering processes . . . . . . . . . . . . . . 88

6.1.3 A typical Raman spectrum . . . . . . . . . . . . . . . . 91

6.1.4 Resonant Raman scattering . . . . . . . . . . . . . . . 92

6.2 Resonant Raman spectroscopy of SWNTs . . . . . . . . . . . 92

6.2.1 Resonant Raman spectra of SWNT . . . . . . . . . . . 93

6.2.2 Radial breathing modes (RBM) of SWNTs - 0 to 350

cm-1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97

6.2.3 The G band of SWNTs - ≈ 1580 cm-1 . . . . . . . . . . 99

6.2.4 The D band of SWNTs - ≈ 1350 cm-1 . . . . . . . . . . 103

6.2.5 The effects of doping on the vibrational modes of SWNTs104

6.2.6 Resonance conditions . . . . . . . . . . . . . . . . . . . 106

6.3 Experimental considerations . . . . . . . . . . . . . . . . . . . 111

6.3.1 Environmental effects upon the resonant

Raman spectra of SWNTs . . . . . . . . . . . . . . . . 112

6.3.2 (i) Effects of contact with the substrate . . . . . . . . . 112

6.3.3 (ii) Thermal effects . . . . . . . . . . . . . . . . . . . . 112

6.3.4 Heating control experiments . . . . . . . . . . . . . . . 116

6.3.5 (iii) Vibrational modes of non-nanotube

components . . . . . . . . . . . . . . . . . . . . . . . . 124

7

6.4 Results and discussion . . . . . . . . . . . . . . . . . . . . . . 129

6.4.1 The SWNT G band . . . . . . . . . . . . . . . . . . . . 129

6.4.2 The SWNT RBM band . . . . . . . . . . . . . . . . . . 142

6.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152

6.5.1 (i) Charge transfer . . . . . . . . . . . . . . . . . . . . 154

6.5.2 (ii) Structural strain . . . . . . . . . . . . . . . . . . . 161

7 Conclusions and future work 163

7.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163

7.2 Future work . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164

Chapter 1

Introduction

1.1 Motivation

Carbon nanotubes are members of the carbon nanomaterials family that

revolutionised the field of Nanoscience and Nanotechnology. After having

been reported in literature for several decades, carbon nanotubes were finally

recognised in 1991 as a new form of carbon [1]. Carbon nanotubes are a one

dimensional material and can be thought of as rolled up sheets of graphene,

a two dimensional carbonaceous material which was only isolated in 2004 [2].

Carbon nanotubes are formed from sp2 hybridized carbon atoms bonded

to form a honeycomb structure. Carbon nanotubes and graphene are very

strong mechanically, due to the strong bonds that form along the sheets

plane. Single-walled carbon nanotubes exhibit semiconducting or metallic

properties depending on how they are rolled [3].

There are two main motivations for functionalizing carbon nanotubes to

produce new hybrid materials, (i) to combine their impressive electronic and

structural properties with those of another system to obtain a new material

with a combination of the properties of both components, and (ii) to make

use of the nanoscale confinement provided by the cavity of the nanotube to

create novel phases of nanomaterials.

1

Single walled carbon nanotubes (SWNTs) can be used to create hybrids in

many ways, either involving the outer surface [4], the inner hollow cavity [5,6]

or even their ends. There are numerous examples of external functionaliza-

tion involving both covalent [4] and non-covalent derivatisation with various

molecular species [7]. The selection of carbon nanotubes as nanoscale con-

tainers for molecular filling is driven by the fact that they provide nanoscale

confinement and form a barrier between the molecular filling and the external

environment [7]; this would be extremely advantageous if the filling material

is for example sensitive to environmental factors such as oxidation. Confine-

ment of material inside carbon nanotubes has been found to be a way to

produce new low-dimensional hybrid nanomaterials with diverse nanoscale

properties and applications [8, 9]. These range from nano-chemistry vessels

[10], atto-gram mass transport [11] and chemical sensors [12], to spin-based

switching for quantum information [13] and vectors for drug delivery [14].

There are two main classes of encapsulated systems: (i) inorganic compounds

in the shape of nanowires and nanocrystals which form directly inside the

nanotubes [6] and (ii) molecular systems which are already pre-formed before

entering the nanotubes [7], [15, 16].

Organic molecules such as porphyrins are very important biologically.

Members of this family are present both in blood (heme) and in plants

(chlorophyll). The strong optical absorption of such organo-metallic molecules

makes them strong candidates for inclusion into photovoltaic devices [17].

Other organo-metallic molecules such as phthalocyanines, a close relative to

the porphyrin family, display useful bulk properties including dichroism and

luminescence, and are used in gas-detection [18]. Organo-metallic molecules

such as metallo-porphyrins are well known as charge donors [19] when intro-

duced to a suitable substrate. The metal ion core of these molecules carry a

non-zero magnetic spin which can make these hybrid systems paramagnetic

and hence interesting for theranostic applications of carbon nanotube hybrids

[14].

In this study single walled carbon nanotubes (SWNTs) have been func-

tionalized both endo- and exohedrally with selected organo-metallic molecules:

NiPc and ClAlPc from the phthalocyanine family and NiTPP from the por-

phyrin family.

2

To our knowledge, hybrids of carbon nanotubes with these molecular

systems have not been produced to date. In general, encapsulation inside

carbon nanotube templates confers a robustness to the hybrid system which

is not present when the molecular systems are attached exohedrally. For this

reason, our primary target was to employ a suitable method to produce en-

dohedral encapsulation. We have chosen a supercritical fluid-based processes

to induce encapsulation inside of nanotubes as this is particularly suitable for

large molecular systems, such as NiTPP, which has low diffusional properties

and therefore cannot be encapsulated by mere thermal diffusion; or for sys-

tems that are air-sensitive and have to be processed in solution. Supercritical

CO2 has not been widely applied as a filling method for carbon nanotubes.

We showed that it produced successful encapsulation in carbon nanotubes

with a range of diameters, from 1.3 up to 3.0 nm. Characterization with

high resolution transmission electron microscopy (HRTEM) revealed previ-

ously unreported row-like ordering of large planar molecules.

Comparisons were also sought from the exohedral functionalization of the

carbon nanotubes with the same organo-metallic systems. This is because

metallo porphyrins and phthalocyanines are expected to functionalize less

effectively the outer surface of the nanotubes (due to chelation with the

central metal ion that perturbs their aromacity). Raman spectroscopy has

been used as the main tool to probe the changes in the electronic properties of

the nanotubes upon both endo and exo-hedral functionalization. The choice

of molecules allowed well-motivated comparisons as (i) NiPc and NiTPP

share the same metal ion core, but differ in their central ring structure and

appendages, (ii) ClAlPc and NiPc share the same body, but have different

central cores which makes the ClAlPc strongly dipolar, and finally (iii) the

three systems have different sizes of the HOMO-LUMO gaps which confers

them different propensity for charge transfer to or from the carbon nanotubes.

3

1.2 Summary of thesis

This thesis has been divided into seven chapters. Chapter 1 introduces

the background of the subject and specific motivation. Relevant elements

of theory for carbon nanotubes are given in Chapter 2. Chapter 3 deals

with the factors that control encapsulation and exohedral functionalization

of carbon nanotubes with guest species. Focus is on the specificity of the

organo-metallic systems used. The suitability of the supercritical CO2 pro-

cess for molecular filling is argued. Chapter 4 describes the instrumentation

developed and implemented concerning the supercritical CO2 processes un-

dertaken in this study. It also outlines the basic purification, characteriza-

tion, washing and functionalization procedures applied to carbon nanotubes.

Chapter 5 describes HRTEM investigations and brings evidence of successful

encapsulation and ordering inside carbon nanotubes. It also presents compar-

isons with relevant encapsulated systems from prior work. Chapter 6 includes

relevant elements of Raman spectroscopy theory and presents comparative

studies applied to the nanotubes endo- and exohedrally functionalized with

the organo-metallic systems. The results are discussed in terms of charge

transfer between the organo-metallic guests and the nanotubes, and induced

structural strain. Conclusions and future work are provided in Chapter 7.

4

Chapter 2

Background and theory

In this chapter the relevant theory for carbon nanotubes is discussed. The

possible types of orbital hybridization of the carbon atom are briefly reviewed

with a focus on the sp2 hybridization found in graphene and carbon nan-

otubes. The lattice structure of graphene is described and the energy disper-

sion relation is obtained using the tight binding method. The structural and

electronic properties of carbon nanotubes are then discussed in terms of those

of graphene.

2.1 Carbon

Carbon is an element which possesses many allotropes. One of the most well

known carbon based structure is diamond. Diamond is both very hard and

a very good electrical insulator. In contrast, another common allotrope of

carbon, graphite, has vastly different properties. It is a very soft material

and a good conductor. It is the hybridisation of the atomic orbitals of carbon

which make such great variation between its allotropes possible.

The carbon atom is the sixth atom in the periodic table. Every atom of

carbon has six electrons, two of which are tightly bound and fully occupy

the spherical 1s orbital.

5

The outer or valence electrons are less tightly bound and occupy the 2s,

2p orbitals. In the ground state, the electronic configuration of carbon is 2s2,

2px1, 2py

1. With just a small excitation it is possible to raise an electron from

the 2s orbital to into a 2pz orbital; this process is called promotion. In the

promoted state the carbon atom has the configuration 2s1, 2px1, 2py

1, 2pz1.

In this state the wavefunctions of the separate atomic orbitals can interfere

to produce hybrid orbitals.

There are three possible hybridization types; sp3, sp2 and sp. In sp3

hybridisation a linear combination of the 2s, 2px, 2py and 2pz results in four

equivalent hybridised orbitals. These form a tetrahedral structure, with each

pointing to one of the four corners of a regular tetrahedron. The overlap of

these hybridised orbitals can form strong molecular bonds called sigma (σ)

bonds. Sigma bonds have cylindrical inter-atomic symmetry. In diamond,

every carbon atom forms four σ bonds with four surrounding carbon atoms.

This gives diamond high 3-dimensional regularity and great strength.

The hybridisation involved in graphite is sp2. The hybridized atomic

orbitals are formed from linear combinations of the 2s, 2px and 2py. The

possible linear combinations are:

h1 = s+ 212py h2 = s+

(32

) 12 px −

(12

) 12 py h3 = s−

(32

) 12 px −

(12

) 12 py

(2.1)

The resulting orbital configuration is trigonal in structure and planar

in nature. The three equivalent hybridised orbitals point to the corners of

an equilateral triangle and are separated by 120o as shown by Figure 2.1

(a). Each of these hybridised orbitals is capable of overlapping with another

hybridized orbital to form a σ bond. The 2pz orbitals do not take part in the

hybridisation process. Instead the two lobes of the 2pz orbital are directed

perpendicular to the trigonal plane out of the plane - this is shown in Figure

2.1 (b).

When two or more carbon atoms with un-bonded 2pz orbitals come into

close proximity, for example in the aromatic ring shown in Figure 2.1 (c), a

molecular bond called a π bond is created.

6

A π bond is formed by the overlap of two in-phase 2pz orbitals.

Figure 2.1: Diagrams of (a) sp2 hybridized σ orbitals of trigonal carbon (adapted from[20]) (b) un-hybridized 2pz orbital (c) schematic diagram of an aromatic ring (d) schematicdiagram showing the delocalized nature of the electrons in the π bonding orbitals of anaromatic ring [21].

In a π bond the probability density of the electron between the atoms

is continuous as shown below in Figure 2.1 (d) and the electrons become

delocalized and are shared between the atoms of the ring. π bonds are

weaker than sigma bonds and as a result the electrons are much less tightly

bound.

2.2 Graphene

Graphene is a planar material of only one carbon atom in thickness. It is

formed from sp2 hybridized carbon atoms arranged in a hexagonal structure.

7

Figure 2.2 shows the structure of graphene. The three sp2 hybridised or-

bitals of the carbon atoms form three σ bonds with the orbitals of three

adjacent carbon atoms. This pattern is repeated throughout the graphene

sheet resulting in its distinctive honeycombed structure.

2.2.1 The graphene lattice

The structure of graphene does not fit into a regular lattice. This is because

it is not possible to create unit translation vectors which will allow one to

reach all of the atoms in the structure. Instead it is necessary to split the

main structure into two sub-lattices. The green dots present in Figure 2.2

represent sub-lattice A and the blue dots represent sub-lattice B. Each of the

sub-lattices is a Bravais lattice and possesses three directions of symmetry.

These are shown as blue dotted lines in the top left part of the figure. Each

line of symmetry is separated by an angle of 120o.

The unit cell of a Bravais lattice is created around a single atom. Due to

the unique structure of the graphene lattice, it is necessary to have one atom

from each sub-lattice in the unit cell. This results in the unit cell of graphene

being a rhombus shape - the unit cells are shown in red. Each of the four

corners of the unit cells lies in the centre of a hexagon. These are the lattice

points of the graphene lattice. This form of unit cell is the smallest shape

that can be fitted to the lattice that includes both atoms of the sub-lattice

and can be repeated to encompass the entire lattice.

8

A B

a1

a2

a1

a2

120o

Figure 2.2: The crystal structure of graphene. Atoms A and B are non-equivalent atoms.The vectors a1 and a2 are the unit translation vectors of the graphene lattice. The redrhombuses are unit cells of the graphene lattice. The blue dotted lines indicate the threedirections of symmetry in the graphene lattice.

The unit vectors of the of the Bravais lattice are given below:

~a1 =(√

3a2, a2

)~a2 =

(√3a2,−a

2

)(2.2)

where |~a1| = |~a2| = a = 2.46 A, and the carbon-carbon sp2 bond length =a√3

= 1.42A.

2.2.2 The reciprocal lattice of graphene

To facilitate the determination of the electronic structure of graphene it is

useful to transform the real lattice into reciprocal space. The reciprocal

lattice of graphene is shown in Figure 2.3 below.

9

b1

b2

K

MK

MΓ

Figure 2.3: The reciprocal lattice of graphene. The unit cells of the reciprocal latticeare shown as red rhombuses. b1 and b2 are the lattice vectors of the reciprocal lattice.The blue hexagon shown in the figure is the first Brillouin zone. Γ, K and M are pointsof high symmetry in the lattice.

Comparing Figures 2.2 and 2.3, it can be seen that the reciprocal lattice is

orthogonal to the real lattice. This is directly related to the general structure

of reciprocal lattice translation vectors.

In general, the reciprocal translation vectors of this two dimensional lat-

tice take the form:

~b1 = 2π(

~a2×~a3~a1·~a2×~a3

)~b2 = 2π

(~a3×~a1~a1·~a2×~a3

)(2.3)

where ~a3= 0i + 0j + ck, ~a1 and ~a2 are the real translation vectors of the

graphene lattice.

Using equations (2.3) the reciprocal lattice vectors of graphene can be

shown to be:

~b1 =(

2π√3a, 2πa

)~b2 =

(2π√3a,−2π

a

), (2.4)

where |~b1| = |~b2| = 4πa√3.

10

The first Brillouin zone is defined as the primitive cell of the reciprocal

lattice. It is useful for describing the electronic properties of a material.

The electronic distribution found in the 1st Brillouin zone is common to the

entire reciprocal lattice. The points labelled Γ, K and M are the points of

high symmetry in the reciprocal lattice of the graphene sheet.

2.2.3 The tight-binding approximation for the energy

band structure

In the idealised view of atoms, electrons occupy well defined orbitals and pos-

sess a quantised energy depending on the orbital type and principle quantum

level. However, when two or more atoms are brought into close proximity, as

is the case in a crystal lattice, the valence electrons interact. The eigenstates

of the valence electrons interfere and create energy bands. In the case of

the graphene lattice the most interesting energy band is formed by the 2pz

valence electrons. These are known as the π bands of graphene. It is these

bands which are closest in energy to the Fermi energy and hence are most

important for charge transport.

In the tight-binding method (sometimes called the linear combination of

atomic orbitals method) the starting point is the set of orbital energy levels of

the tightly bound electrons of a single isolated atom. With other near neigh-

bour atoms brought near to this originally isolated atom, the wavefunction of

the original atom would be overlapped by those of the near neighbour atoms.

If the atoms are far enough apart, this overlap would be small enough for it

to be taken into account just by making comparatively small corrections to

the originally isolated atom model. The overall picture of the energy levels

of a lattice of atoms is made up of slightly modified atom models [22].

The tight-binding method can be used to calculate an empirical solution

for the band structure of graphene as follows [3].

11

The crystal lattice of graphene possesses translational symmetry along the

lattice vectors ~a1 and ~a2, this implies that any wavefunction of the lattice,

Ψ, should obey Blochs theorem

T~aiΨ = ei~k·~aiΨ (2.5)

where in graphene i = 1, 2 and T~ai is a lattice translation operator and ~k is

the wave vector of the lattice.

The wavefunctions that satisfy this condition are called Bloch orbitals.

There are two Bloch orbitals for graphene, one for each sub lattice. These

take the form shown below.

Φj(~k, ~r) =1√3

N∑~Rj

ei~k·~Rjφj(~r − ~Rj) (2.6)

where (j = A,B) and ~Rj and ~r are position vectors of the atom j and the

2pz orbital of atom j respectively. The atomic orbital of atom j is designated

by φj(~r − ~Rj). With each individual orbital multiplied by its phase factor

ei~k·~r, a summation over the whole lattice (N unit cells) will produce Φj(~k, ~r).

The wavefunction of the crystal lattice, Ψ(~k, ~r), is a linear combination of

ΦA and ΦB.

Ψj(~k, ~r) =∑

j,j′=A,B

Cjj′Φj′(~k, ~r), (2.7)

where j, j′ = A,B and Cjj′ are complex coefficients.

The energy eigenvalues E(~k) of the eigenstate (2.7) can be found by

solving the time independent Schrodinger equation

H|Ψj〉 = E(~k)|Ψj〉 (2.8)

where H is the Hamiltonian of the graphene lattice.

12

By taking the inner product of (2.8) and inserting (2.7), it is possible to

derive the secular equation.

The energy dispersion relation of graphene is described by the following

equation:

det[H − ES] = 0 (2.9)

In this case, Hjj′ = 〈Φj|H|Φj′〉 and Sjj′ = 〈Φj||Φj′〉 are the transfer

integral matrices and overlap integral matrices respectively.

The secular equation for graphene is given by

∣∣∣∣∣HAA(~k − E(~k) · SAA(~k) HAB(~k − E(~k) · SAB(~k)

HBA(~k − E(~k) · SBA(~k) HBB(~k − E(~k) · SBB(~k)

∣∣∣∣∣ = 0 (2.10)

In order to solve the secular equation, it is necessary to take the lattice

structure of graphene into account.

The three nearest neighbours belong to a different sub-lattice (surround-

ing atoms belong to the blue sub-lattice - see Figure 2.4). With these con-

siderations in mind a number of simplifications can be made to the secular

equation. As we are only considering interactions between nearest neigh-

bours (shown schematically in Figure 2.4), it is only necessary to integrate

over the single atom in HAA and HBB. This results in a number of useful

simplifications,HAA = HBB = ε2pz , SAA = SBB = 1 and SAB = sf(~k) = S∗BA. Here ε2pz is the energy of the 2pz orbital, s is the nearest neighbour overlap

integral and f(~k) = (ei~k·~R1 + ei

~k·~R2 + ei~k·~R3).

The contributions from the three nearest neighbours in the lattice are

described by

HAB = HBA = γ0 · (ei~k·~R1 + ei

~k·~R2 + ei~k·~R3) = γ0 · f(~k) (2.11)

13

R1

R2

R3

Figure 2.4: The three nearest neighbour atoms relative to atom are shown as red dots.The position vectors ~R1, ~R2 and ~R3 give the relative locations of the three nearest neigh-bours.

where γ0 is the nearest neighbour transfer integral

γ0 = 〈ΦA(~r − ~RA)|H|ΦA(~r − ~RA − ~R1)〉 (2.12)

Inserting the above simplifications into the secular equation allows for

eigenvalues E(~k) to be found. Here E(~k) is a function of w(~k), kx and ky:

E(~k) =ε2pz ± γ0w(~k)

1± sw(~k)(2.13)

The positive and negative signs combine to give either the bonding energy

band π (positive combination) or the anti-bonding π* band (negative). The

ε2pz and s terms are constant which are important in determining the absolute

energy but not for appreciating the form of the energy bands.

14

The function w(~k) is given by:

w(~k) =

√|f(~k)|2 =

√1 + 4cos

√3kxa

2cos

kya

2+ 4cos2

kya

2(2.14)

where kx and ky are lattice wave vectors of the reciprocal lattice respectively

and a is the modulus of the real lattice translation vectors [23].

2.2.4 The electronic band structure of graphene

The first Brillouin zone of the reciprocal lattice contains all of the points

of high symmetry necessary to understand the unique band structure of

graphene - these are shown in Figure 2.5 below.

Γ

K

M

K'

Figure 2.5: Expanded view of the first Brillouin zone of graphene labelled with the pointsof high symmetry (red circles), Γ, K, M and K ′. K and K ′ are non-equivalent points asthey correspond to the two non-equivalent A and B sub-lattices of the direct lattice.

The non-equivalent K and K ′ points, present at the corners of the Bril-

louin zone, mirror the atomic symmetry of the real lattice. Plotting E(~k)

along the lines of high symmetry allows for a graph of energy as a function

of wave vector to be plotted, this is shown in Figure 2.6 below.

15

π*

π

Wave vector

En

erg

y (

eV

)

M M K Г

0

6

12

-12

-6

Fermi energy

Figure 2.6: Plots of E(~k) as a function of ~k along the lines of high symmetry, M → Γ,Γ → K and K → M. The plot follows the perimeter of the triangle in k space shown inFigure 2.5. The red dashed square is centred on one of the Dirac points. Adapted from[21].

The two energy bands closest to the Fermi level are the π bonding and the

π∗ anti-bonding bands respectively. These are highlighted in blue and red in

the energy band diagram of Figure 2.6. Each of the two atoms in the lattice

unit cell contribute one electron to the π bonds, these fully fill the π bonding

energy band. It can be seen from the figure that the valence band (occupied π

band) and the conduction band (unoccupied π∗ band) meet at the K points;

this classifies graphene as zero band-gap semiconductor material.

The K and K ′ points located at the 6 corners of the first Brillouin zone are

known as Dirac points. As the band energy approaches the Fermi level, the

electronic density of states at the Dirac points tends to zero. An interesting

feature of the band structure of graphene is the shape of the conduction and

valence bands above and below the Fermi level. It can be seen from Figure

2.7 that both bands are cone shaped at the Dirac points. The σ bands of

graphene have been excluded from the figure as they are of much greater

energy (≈ 14eV [3]) than the π bands and therefore do not take any part in

conduction, which involves energy levels around the Fermi level.

16

Figure 2.7: π energy bands, E, as a function of wave vector (~k). Insert diagram showingthe linear nature of the π bands close to the π Dirac points [21].

2.3 Carbon nanotubes

2.3.1 Structure

Carbon nanotubes are a unique allotrope of carbon. Nanotubes can be

thought of as being rolled sheets of graphene. The carbon atoms in a nan-

otube arrange themselves in hexagonal cells which link together to form long

cylinders of up to 1 µm in length and diameters ranging from 0.5 to 10nm

in diameter. A nanotube formed from a single sheet of graphene is known as

a Single Walled Carbon Nanotube (SWNT), an example of a single walled

armchair type nanotube is shown in Figure 2.8. The large majority of this

work will involve SWNTs.

Graphene sheet

rolled up

Figure 2.8: The rolling of a single walled carbon nanotube of the armchair type (basedupon [24]).

17

It is also possible to have Multi-Walled Carbon Nanotubes (MWNTs).

These are made from a number of concentric single walled nanotubes of

increasing diameter, as show in Figure 2.9.

Figure 2.9: Schematic diagram of a 4 walled MWNT (end on).

There are many ways in which a graphene sheet can be rolled to form a

carbon nanotube, with a wide range of resulting diameters and chiralities.

The chirality of a nanotube describes the direction in which the graphene

sheet has been rolled to create it. Every nanotube has an associated chiral

vector ~C which forms the circumference of the nanotube.

The chiral vector determines the structural type and the electronic prop-

erties of the nanotube. The general form of a chiral vector is

~C = n~a1 +m~a2 = (n,m), (2.15)

where n and m are integers [3].

There are three possible nanotube classes, armchair, zig-zag and chiral,

these are shown in Figure 2.10.

Each chiral vector has an associated chiral angle, θ. This angle can be

determined using the following equation [3]:

cosθ =~a1 · ~C|~a1| · |~C|

=2n+m

2√n2 + nm+m2

(2.16)

18

x

y

Graphene sheet

rolled from this

edge

a1

a2

θ

Zig-zag vector

Chiral vector

Armchair vector

(4, 2)

(3, 3)

(0, 5)

Tra

nsla

tion v

ecto

r

Figure 2.10: The possible chiralities of carbon nanotubes. The chiral and translationvectors of a (4,2) chiral nanotube are shown as red and blue lines respectively - the chiraland translation vectors are perpendicular to one another. The chiral angle, θ, is measuredrelative to the zig-zag line. The chiral vectors of (0, 5) zig-zag and (3, 3) armchair nan-otubes have also been included on the diagram in magenta and green respectively. Thegreen rectangle represents the unit cell of a (3, 3) armchair nanotube.

The chiral angles and structures of each type of nanotube are shown in

Figure 2.11.

19

Figure 2.11: The different types of nanotube chirality and the corresponding chiral angle(Adapted from [25])

Given that the chiral vector forms the circumference of the nanotube, it is

possible to determine the diameter, d, of the nanotube directly by the using

the following equation [3]:

d =|~C|π

=a

π

√n2 + nm+m2 (2.17)

The last structural property of nanotubes to be described is the trans-

lational symmetry along the length of the tube. This is determined by the

translation vector,

~T = t1~a1 + t2~a2 (2.18)

where t1 and t2 are integers which can be determined from the chiral indices

n and m using the equations t1 = 2m+ndR

and t2 = −2n+mdR

where dR is the

greatest common divisor of (2m+ n) and (2n+m).

20

The translation vector, ~T , and the chiral vector, ~C, form a rectangle on

the graphene lattice [3]. This is the unit cell of the nanotube, an example of

the unit cell of a ~C = (3, 3) arm chair nanotube is shown as a green rectangle

in Figure 2.10. The size of the minimum translation vector and hence the

unit cell depends upon the chirality of the nanotube. For armchair nanotubes

|~T | = a and |~C| = n√

3a, while for zig-zag nanotubes |~T |=√

3a and |~C| = na

[3]. Chiral nanotubes can have relatively large unit cells with the maximum

occurring at an angle of 15o. Calculating the number of carbon atoms in

a given nanotube unit cell is a reasonably simple procedure. Firstly it is

necessary to calculate the number of hexagons, q, present in the unit cell of

the nanotube. This can be achieved by dividing the area of the nanotubes

unit cell, St = |~T × ~C|, by the area of one hexagonal cell of graphene,

Sg = |~a1 × ~a2|,

q =stsg. (2.19)

Given that there are two carbon atoms per hexagon, the total number of

carbon atoms in the nanotube unit cell, nc, is given by nc = 2q (based upon

[3]).

2.3.2 The reciprocal lattice of SWNTs

In order to effectively describe the electronic structure of carbon nanotubes

it is necessary to construct the first Brillouin zone of the reciprocal lat-

tice. Firstly, considering the lattice translation vectors along the length and

around the circumference of the tube. The reciprocal lattice translation vec-

tor in the z direction of the tube, ~kz, is given by:

~kz =2π

~T(2.20)

Since the diameter of a nanotube is very much less than its length, it

can be thought of as being infinitely long. This results in a continuous wave

vector ~kz along the tube.

21

As a result the first Brillouin zone extends from ~kz = −πa

to πa

. The

wavevectors in the circumferential direction, ~k⊥ , are more interesting. Any

wave vector, ~k⊥, is quantised according to the following boundary conditions

µ · λ = |~C| = π · d. Wave vectors which satisfy this take the form:

~k⊥,µ =2π

λ= µ · 2π

~C(2.21)

where, λ is the wavelength of the electron wavefunction and µ is the quan-

tisation number which is an integer and can take any value from - q2+1, 0,

1 ...., q2. For armchair and zig-zag nanotubes q = 2n, while for chiral nan-

otubes q =2(n2

1+nm+n22)

dcdRcd. Here dcd is the greatest common divisor of (n,m)

and Rcd = 3 if (n−m)/3n is an integer and Rcd = 1 otherwise [23].

This can be explained by considering the behaviour of an electron present

on the circumference of the nanotube. An electron can only exist when the

corresponding wavefunction meets the above boundary conditions. Most

importantly, the wavefunction must possess a phase shift which is an integer

multiple of 2π. All other phases result in destructive interference.

We now have the necessary wave vectors to describe the first Brillouin

zone of a carbon nanotube. The first Brillouin zone consists of a rectangular

band of parallel lines of length 2πa

(~kz) directed along the ~kz axis. These

lines are quantised in the circumferential direction by the spacing of 2π

| ~C|. A

detailed diagram of the first Brillouin zone of a (3,3) armchair nanotube is

given in Figure 2.12.

22

Figure 2.12: The first Brillouin zone of a (3,3) armchair nanotube.

For a given type of nanotube, the size of the Brillouin zone in the re-

ciprocal lattice direction (~kz) is constant. As the diameter of the nanotube

increases, the number of quantised k lines around the circumference also in-

creases. However, an increase in diameter also results in a decrease in the

separation between the k lines. As a consequence of this, the electronic

quantisation becomes less distinct with increasing diameter.

2.3.3 Electronic structure of SWNTs

If the k lines of a nanotube match up with the K points of the graphene

Brillouin zone then the nanotube will be metallic. The nanotube shown in

Figure 2.12 is metallic as are all armchair nanotubes; if the k lines do not

intersect with any of the K points the nanotube will be semi-conducting.

The zig-zag nanotubes can be either metallic or semiconducting; in addition

chiral nanotubes can be a mixture of metallic and semiconducting. The basic

principle of the zone folding approximation is that the electronic band struc-

ture of a nanotube can be determined from the band structure of graphene

along the allowed k lines.

23

With the zone folding method, the band structure of graphene can be

adapted to give the band structure of a nanotube by replacing the wavevec-

tors of the graphene lattice by those of the nanotube (i.e. kx → k⊥, ky → kz).

The band structure of a generic (n,0) zig-zag nanotube is described by the

following equation:

Eµ = ±t

[1± 4cos

(µ · πn

)cos

(√3ka

2

)+ 4cos2

(µ · πn

)] 12

(2.22)

where Eµ is the dispersion energy of the nanotube energy bands, t is the

transfer integral, − π√3< ka < π√

3and n is a chiral integer and [3].

Similar expressions can be derived for armchair and chiral nanotubes.

Figure 2.13 shows a plot of expression (2.22):

E (eV)

Γ k

μ = 0

μ = 1 and -1

μ = 2 and -2

μ = 3 and -3

μ = 4 and -4

μ = 5 and -5

μ = 6 and -6

μ = 7 and -7

μ = 8

μ = 8

μ = 7 and -7

μ = 6 and -6

μ = 5 and -5

μ = 4 and -4μ = 3 and -3

μ = 2 and -2

μ = 1 and -1

μ = 0

8

6

4

2

-8

-6

-4

-2

0.2 0.4 0.6-0.2-0.4-0.6

Figure 2.13: Band structure of an (8,0) semiconducting zig-zag SWNT from zone foldingof graphene band structure. Image based upon [21].

The lowest energy bands of the first Brillouin zones for metallic and semi-

conducting nanotubes are shown in Figure 2.14.

24

Figure 2.14: The band structures for (a) a metallic nanotube and (b) a semiconductingnanotube.

2.3.4 Electronic density of states

The electronic density of states is defined as the number of electronic states

available in a given energy range; it is a quantity that is both useful for de-

scribing the electronic structure of a given nanotube, and is also measureable.

The density of states is dependent upon the dimensionality of an object

- a carbon nanotube is classed as a 1 dimensional object (z direction) and is

quantised in 2 directions i.e. x and y. The density of states, in a transverse

direction, for a one dimensional object varies as 1/√E , where E is the

transverse energy. It can be shown [26] that the density of states for the first

Brillouin zone of a nanotube is described by:

n(E) =4a

π2dγ0

∞∑µ=−∞

g(E,Eµ), (2.23)

with:

g(E,Eµ) =

{|E|/

√E2 − E2

µ |E| > |Eµ|0 |E| < |Eµ|

(2.24)

25

The density of states for an (8,0) semiconducting zig-zag SWNT and a

(5,5) metallic armchair SWNT are shown in Figure 2.15.

E11

E11

-10.0 -5.0 0.0 5.0 10.0

Energy (eV)

(a)

(b)

Figure 2.15: Density of states (DOS) as a function of energy, for a selection of SWNTs:(a) semiconducting and (b) metallic. The transition E11 has been labelled in red as anexample (based on [27]). It is worth noting the differences between semiconducting andmetallic nanotubes at the Fermi energy (Energy = 0), specifically, in semiconductingnanotubes the DOS is zero - in contrast for metallic nanotubes it is non-zero.

The following points are of note (Figure 2.15):

1) At the quantisation energies E = Eµ the function g(E, Eµ) diverges

and produces what is called a van Hove singularity [3].

2) For the special case of Eµ = 0, g(E,Eµ) becomes unity, the conduction

and valence bands of the nanotube meet at the Fermi energy - the nanotube

is metallic (Figure 2.15. b)).

26

The density of states peak at the singularities, as a result they dominate

the physical properties of the nanotube. Concerning the optical properties of

nanotubes, for a photon to be absorbed it has to match the energy interval

Eii, (see Figure 2.15) between symmetric singularities in the valence and

conduction bands [27].

27

Chapter 3

The encapsulation of molecular

systems by carbon nanotubes

using a supercritical carbon

dioxide-based method

In this chapter the factors affecting the binding of molecules to the external

and internal surfaces of a carbon nanotube are reviewed. We targeted the

encapsulation of organo-metallic molecular systems via a supercritical CO2

process, whose physical mechanisms and advantages are described in detail.

The choice of the molecular systems to be encapsulated is also justified: it

includes molecules that are sensitive to air (and would benefit from having the

nanotube protection), molecules that have large dimensions and hence lower

diffusion towards encapsulation, and aromatic planar molecules with metallic

cores which make them more difficult to create hybrids with nanotubes by

exohedral functionalization.

28

3.1 The effect of curvature upon the

reactivity and binding capability of SWNTs

It is well known that graphite is formed from a large number of individual

graphene layers stacked one on top of each other. The graphene layers are

held together by weak van der Waals interactions between the individual

sheets. It is the π bands of graphene which are responsible for this inter-

action and it is known as π-π stacking [28]. In an ideal sheet of graphene

the reactivity of the upper and lower surfaces are equivalent. In carbon nan-

otubes, however, the situation is more complicated because the curvature

of the nanotube side-walls has a significant effect upon the reactivity of the

nanotube.

Two parameters directly linked to the intrinsic curvature of nanotube

side-walls and hence to their reactivity are the pyramidalization angle (θP)

and the π-misalignment angle (φ) [29].

Geometrically speaking, carbon nanotubes are significantly different from

graphene. The sp2-hybridized carbon atoms in an ideal graphene sheet are all

perfectly trigonal, that is, the angle between the π orbitals and the σ bonds is

exactly 90o. In contrast, in carbon nanotubes the curvature of the nanotube

side-walls causes the natural trigonal geometry to become distorted. The

extent of the distortion can be understood in terms of the pyramidalization

angle, which is defined as follows:

θP = (θσπ − 90o), (3.1)

where θσπ is the curvature induced deviation angle between the upper π

orbital and the σ orbitals [29]. The pyramidalization angles of graphene and

a generic SWNT are compared in Figure 3.1 (a) and (b) respectively.

29

Figure 3.1: Diagrams of the pyramidalization angle θ of (a) graphene and (b) of ageneric nanotube, respectively; the π misalignment angle φ of (c) graphene and (d) ageneric nanotube respectively (adapted from [29]).

Another distortion introduced by the curvature of the nanotube is the mis-

alignment of the π-orbitals between adjacent carbon atoms in the nanotube

side-walls, this mis-alignment is only present in carbon atoms which have a

C-C bond that is not parallel to the circumference of the nanotube.

30

In the planar geometry of graphene the π-orbitals between the carbon

atoms are parallel. However, in a nanotube, the curvature of the walls is

such that this is not always the case and an orbital mis-alignment angle φ

develops. A comparison between the mis-alignment angles of graphene and

a generic nanotube is given in Figure 3.1 (c) and (d).

The pyramidalization and misalignmentof the π-orbitals of the C atoms

in the sidewalls of the SWNTs induces local strain and as such the exterior

surface of carbon nanotubes is expected to be more chemically reactive than

the surface of a graphene sheet [29]. Both of these effects, and hence the

local strain in the nanotube, are inversely proportional to the diameter of

the nanotube [29].

The exterior of SWNTs is also expected to be more reactive than the

interior. This is because the pyramidalization of the nanotubes C atoms

causes the exohedral lobes of the π-orbitals to be larger than those of the

interior [30]. The larger lobes on the exterior favour overlap with chemically

suitable atoms or molecules - the reduction in local strain provided by such

an interaction also plays a part [30]. The difference between the reactivities

of the exterior and interior increases with increasing pyramidalization angle

and decreases with increasing diameter. However, the differences are only

moderate for typical examples such as (10, 10) SWNTs with diameters of 1.4

nm [30].

3.2 Effects of curvature upon molecular

adhesion

The curvature of the sidewalls of SWNTs make the interior and exterior

surfaces different geometrically. The nanotube sidewalls curve away from an

adatom / molecule attached externally and towards an adatom/ molecule

attached internally. This has a strong effect on placement of adatoms /

molecules adsorbed onto the internal and external surfaces of nanotubes.

31

The simplest model which gives qualitatively correct results for specific

cases of adatom-nanotube binding energy is the empirical Leonard-Jones po-

tential (LJP) which describes dispersive interactions between neutral atoms.

This empirical model has been successfully applied to a number of graphitic

materials. For a C60 molecule interacting with CNTs, the binding energy

was found to be dependent upon molecule placement [31].

For C60 molecules interacting with nanotubes, the binding energy to the

interior surfaces of the nanotube was found to be greater than when at-

tached to the exterior or to the open ends of the nanotube [31]. The greatest

binding energy was found for C60 molecules interacting internally with the

hemispherical end caps of a nanotube [31]. This can be understood in terms

of geometry matching - the closer the geometric match between the molecule

and nanotube surface, the greater the number of atoms available to con-

tribute strongly to the binding potential [32] - this is shown schematically in

Figure 3.2.

Adatom

Nanotube

Carbon atom of

Nanotube

ra

rb

ra > rb

Figure 3.2: Schematic diagram of the geometric match between adatoms attached to theexterior and interior walls of a nanotube. ra and rb represent the distances between thecenters of the adatoms and the carbon atoms in the walls of the nanotube.

32

To determine the binding energy between a molecule and nanotube, the

binding energy between the molecule and the carbon atoms of the nanotube

are summed; the potential energy of interactions between the molecule and a

single carbon atom in the nanotube,V LJ is described by the empirical relation

V LJ = 4ε

[(σr

)12−(σr

)6], (3.2)

where r is the distance between the molecule and the individual carbon atom,

ε is the depth of the potential well and σ is the hard-sphere radius of the

molecule [5]. For example, calculations using a LJP model show that in

interactions between adatoms and nanotube bundles the relative size of nan-

otube diameter and the van der Waals radius of the adatoms is critical in

determining uptake of adatoms from the gas phase. As one would expect,

small enough molecules can easily fit inside both the nanotubes and in the

interstitial channels between tubes. There will be limits depending upon in-

dividual adatom species and nanotube diameter beyond which adatoms will

not fit inside either space [33].

To reveal qualitative behaviour of an adatom interacting with a nanotube,

a simplified model using the LJP can be used where an arbitrary adatom of

van der Waals diameter SvdW interacts via van der Waals interactions with

an arbitrary nanotube bundle made of SWNTs, the diameter of a SWNT

being d.

Varying the diameter of the nanotube, has a marked effect on the binding

energy between the atom or small molecule and the SWNT upon which it

is adsorbed. As d → ∞ (Figure 3.3 (a)), the interaction strength between

the adsorbed adatom and the SWNT becomes equivalent for the interior and

exterior surfaces. The binding energy for an adatom to the walls of an infinite

diameter nanotube is equivalent to that of graphene. As d is decreased the

binding energy between the adatom and the exterior of the SWNT decreases.

33

Conversely, the binding energy between the adatom and the interior sur-

face of the nanotube increases as d is decreased and reaches a maximum bind-

ing energy at an ideal diameter (doptimum) specific to each atom or molecule

- see Figure 3.3 (b). The ideal diameter for an arbitrary adatom is described

by the following relation:

doptimum = SvdW + (2× rvdW), (3.3)

where rvdW = 0.15 nm and is the thickness of the nanotubes π-orbitals [34];

SvdW is the hard-sphere radius of the adatom. At the ideal diameter, the

binding energy strongly favours the internal site [5]. At larger diameters

Figure 3.3 (c) the interior is still favoured but not as strongly as at the ideal

diameter, this is due to the differences in size. As d is decreased below doptimum

(Figure 3.3 (d)), the repulsive 1/d12 component of V LJ becomes dominant

due the overlap of the π-orbitals of the nanotube and those of the adatom,

making encapsulation unfavourable.

From this simple model it is clear that the van der Waals diameter of

the atom or molecule relative to the diameter of the nanotube is the most

important parameter associated with the placement of the atom or molecule

[5].

The molecules used in this study are shown in Figure 3.4 and the effective

hard-core diameters (SvdW ) of the molecules and the corresponding ideal

nanotube diameters doptimum calculated using equation (3.3) are shown in

Table 3.1.

34

rvdW

SvdWd

d = doptimum

(b)

d >> SvdW

(c)

SvdW

d

rvdW

SvdW d

d < SvdW

(d)

d

(a)

rvdW

∞

rvdW

Adatom

Graphene

Nanotube

Figure 3.3: Schematic diagrams of adatoms attached to carbon nanotubes with (a) d →∞ (b) d = doptimum (c) d >> SvdW and (d) d < SvdW . All symbols as defined in the text.

Cobalt carbonyl (Co2(CO)8) is a roughly spherical molecule with an effec-

tive hard-sphere diameter of 0.5 nm and a corresponding optimum nanotube

diameter of 0.8 nm. The spherical geometry should give this molecule a high

interaction energy with the inner walls of the nanotubes. The cobalt carbonyl

molecule has a core formed from two paramagnetic cobalt atoms.

35

(a)

Cobalt atom

Carbon atom

Oxygen atom

SvdWb

(b)

Nickel ion

core

Phthalocyanine body

(d)

Phenyl group

Nickel ion

core

Porphyrin body

(c)

Chloro-aluminium

dipole core

Phthalocyanine body

SvdWc SvdWd

Figure 3.4: Schematic diagrams of the molecular strutcure of (a) Cobalt carbonyl (Co2(CO)8) (based upon [35]) (b) nickel phthalocyanine (NiPc) (based upon the structureof CoPc [36]), (c) chloro-aluminium phthalocyanine (ClAlPc) and (d) nickel tetraphenylporphyrin (NiTPP) (based upon [37]). The green rings around the square TPP and Pcmolecules represent the van der Waals diameters of the molecules.

In contrast, the other molecules used in this study are significantly larger,

planar in nature and roughly square shaped. In order to obtain an effective

hard-core diameter for these molecules, a circular symmetry has been as-

sumed - see Figure 3.4. Both of the phthalocyanine (Pc) molecules, chloroalu-

minium Phthalocyanine (ClAlPc) and nickel Phthalocyanine (NiPc) molecules

share the same structure and hence have the same effective hard-core diame-

ter of 1.5 nm with an associated ideal nanotube diameter of 1.8 nm. However,

they have dissimilar cores: NiPc has a paramagnetic nickel ion at its core,

while ClAlPc has a Cl-Al dipole, which forms an electric dipole. The nickel

tetra phenyl porphyrin (NiTPP) molecule is the largest of the set, possessing

an effective hard-core diameter of 2 nm and an associated ideal nanotube

diameter of 2.3 nm. It has a nickel ion at its core, as with the NiPc molecule,

however, NiTPP has a porphyrin body with phenyl appendages attached.

36

Molecular species SvdW / nm doptimum / nm

Co2(CO)8 0.5 0.8

ClAlPc 1.5 1.8

NiPc 1.5 1.8

NiTPP 2.0 2.3

Table 3.1: The optimum nanotube filling diameters doptimum for the molecules used inthis study, SvdW is the size of the molecules inluding the van der Waals radii of theconstituent atoms. The sizes of the van der Waals potential surface of the Pc moleculeswere based upon that of cobalt phthalocyanine (CoPc) as quoted by [36]. Those of theNiTPP and Co2(CO)8 molecules were determined from molecular models.

Taking symmetry into account, the LJP interaction between these molecules

and the inner surfaces of the nanotubes would be maximised if the molecules

were arranged cross-ways along the axis of the nanotube - Figure 3.5 shows

this schematically using the NiTPP molecule as an example. Given the

square geomtery of the molecules they are likely to have a weaker LJP inter-

action than the cobalt carbonyl interaction - this would result from greater

symmetry mis-match. However, this is not the geometry observed upon en-

capsulation (see chapter 5 section 5.4).

SWNT

NiTPP molecule

Figure 3.5: Schematic diagram showing a NiTPP molecule positioned in a SWNT ar-ranged cross-ways to the length of the nanotube (based upon [38]) The yellow outlinerepresents the van der Waals surfaces of the molecule and nanotube.

37

3.3 Aromatic interaction between molecules

and the exterior of CNTs

For aromatic molecules it is possible for intermolecular interactions to occur

through π stacking. In π stacking, the planar π systems of aromatic molecules

lay one on top of each other, in a nearly parallel orientation [20].

The cobalt carbonyl molecule is not aromatic and as such cannot interact

in this way. The other molecules of this study, however, posses multiple

aromatic sub-structures which could enable π stacking to occur with the

nanotubes exterior as well as between themselves. The aromatic molecules

are shown in Figure 3.6 (a) and (b) with the structures likely to result in π

stacking highlighted.

The TPP and Pc molecules have two possible sub-structures where π

stacking between the molecule and surfaces of the nanotube could occur:

(i) the central ring systems

(ii) the benzene-like appendages attached to the central ring systems

Taking the porphyrin molecule as an example, the possible binding areas

for the sub-structures on a nanotube of 2.3 nm in diameter is shown in

Figure 3.6 (c) and (d). The porphyrin is also shown encapsulated inside of

the nanotube, however, given that the molecule is separated from the inner

walls of the nanotube, it is unlikely that π stacking would occur - unless

the molecule were to be distorted to more closely match the geometry of the

nanotube.

38

SWNT

Outer van der Waals

π surface of nanotube

NiTPP molecule

Outer van der Waals

π surface of NiTPP

molecule

Outer van der Waals

π surfaces of aromatic

rings

Inner van der Waals

π surface of nanotube

TPP molecule

(a)

Pc molecule

(b)

(c) (d)

Figure 3.6: Schematic diagram showing the (a) Pc and (b) TPP molecules respectively.The structures likely to be involved in π stacking are highlighted in red, appendages andgreen, central ring system. (c) and (d) show the possible binding areas on a nanotube fora fully aromatic porphyrin molecule (c) and (d) where the aromacity of the central ringhas been destroyed by chelation with a metal ion.

It has been found that un-chelated porphyrins (porphyrins without a

metal centre), for example tetraphenyl porphyrin (TPP) shown in Figure

3.6 (b), show strong affinities for the exterior of semiconducting CNTs, so

much so that the union of the two can be used to separate semiconducting

nanotubes from a metallic/ semiconducting mixture [39]. However, upon

chelation with a metal ion complex, the binding between metallo-Porphyrins

and nanotubes becomes less favourable [39]. This highlights that the primary

centre for the binding of the un-chelated TPP to the SWNT is the porphyrin

ring system (Figure 3.6 (c)) and not the phenyl groups.

39

This also seems to be true for Phthalocyanine molecules in which upon

the formation of a coordination bond between a metal ion and the nitrogen

atoms, aromaticity of the central ring system is destroyed and thus π stacking

via it is prevented [40]. In conclusion, the attachment of either the NiTPP,

NiPc or ClAlPc molecules to CNTs via π stacking involving the centres of

these systems seems unlikely.

The binding energy of aromatic molecules adsorbed to the exterior of

CNTs by π stacking has been found to be weak for benzene-like molecules.

However, if charge transfer between the nanotube and molecule occurs, the

binding energy can be significantly higher [41]. It is well established that met-

alloporphyrins are electron donating systems [19], therefore this may present

a route through which π stacking could occur between the appendages of

the molecules and graphitic surface of the CNT (Figure 3.6 (d)). We fi-

nally note that in reference [42] CoOEP molecules were found to exohedrally

functionalize SWNTs in high yield despite the presence of the metallic core.

3.4 Effects of curvature upon molecule

diffusion on and in SWNTs

Calculations of the potential energy and molecular dynamics with a Leonard-

Jones type potential found that both the curvature and helicity of the a car-

bon nanotube have a great bearing on the diffusion of an adatom adsorbed

onto the nanotube walls [43]. It was found that adatom diffusion is very de-

pendent upon the curvature of the nanotube. The diffusion barrier increases

monotonically with curvature [43].

Therefore it is possible to conclude that diffusion along the interior of the

nanotube (negative curvature) is easier than along graphite (zero curvature),

however diffusion along a flat surface is easier than along the exterior of the

nanotube (positive curvature). This can be understood in terms of curvature

induced strain in the walls of the nanotube, the inner surface experiencing a

negative strain (compression) and the exterior experiencing positive strain.

40

The helicity has also been shown to impact upon the diffusion path of

an adsorbed adatom, with diffusion being favoured in armchair rather than

zig-zag nanotubes [43]. With these considerations in mind it is reasonable

to conclude that the interior of the nanotube is a more suitable environment

for the formation of ordered nanostructures than the exterior.

Calculations have shown that due to a smoother potential energy surface

in the interior of nanotubes, the barrier to diffusion is lower than that of

the exterior surface even though the binding energy endohedrally is greater

[43]. This makes the diffusion of adatoms through the interior of the nan-

otube much more favourable than along the exterior. Once the adatoms are

inside, for adatoms of or near optimum size for the nanotube, the increase

binding energy inside makes the removal of encapsulated adatoms demanding

energetically.

3.5 Molecular encapsulation by SWNTs

3.5.1 Encapsulation methods

For filling carbon nanotubes with molecular species more traditional methods

used either thermal diffusion [44] or filling from solution [5, 15,45,46].

Thermal diffusion of C60 fullerenes has been shown to produce SWNT

filling yields of up to 85%. However, to achieve this result, a temperature

of 650o C was required [44]. The decomposition temperatures of a lot of

organic molecules of similar size to C60 are far below this, indeed the cobalt

carbonyl molecule is known to decompose above a temperature of 52o C

[47]. Therefore, the high temperatures necessary for this process to occur

make it especially unsuitable for the encapsulation of molecules such as this.

Porphyrins however, have a much greater thermal stability and enter the

molten phase at temperature of 200 - 300o C [48]. Results indicate that

porphyrin peapods can be formed using a thermal diffusion method with a

process temperature of 400o C. [44]. Phthalocyanine molecules also possess

a high degree of thermal stability - for example, the NiPc molecule is known

to be stable in Ar up to a temperature of 750o C [49].

41

Therefore, a thermal diffusion method may be suitable for the porphyrin

and phthalocyanine molecules but might not be compatible with the cobalt

carbonyl molecule. Nevertheless, the larger the molecule (e.g. the NiTPP

molecule used here), the more difficult its encapsulation by thermal diffusion

[5].

A number of studies have been conducted upon filling from solution using

a variety of organic solvents, both aromatic, such as toluene, and polar, such

as ethanol - these met with mixed results [5,15,45,46]. The main advantage

of filling by solution is that it is a relatively low-temperature process and

therefore suitable for thermally unstable molecules.

In order to maximise the yield of encapsulated molecules upon filling from

solution there are a number of criteria which must be optimised [5]:

(i) The solvent used must have a surface tension (γ) which is low enough

to wet the surfaces of the CNT.

(ii) The interaction between the solvent and the solute (molecules to be

encapsulated) must be weak.

(iii) The interaction between the solvent and the CNT must also be weak.

(iv) The interaction between the CNT and the solute must be relatively

strong.

(v) The critical diameter of the solvent molecules should be small enough

for them to escape from any encapsulated structure which has formed

using the filling process.

It has been found that a solution’s ability to effectively wet the surface

of the nanotube is critical to whether it will induce molecular encapsulation

[50]. The parameter which has the greatest effect on a solvent’s ability to wet

the surface of a CNT is the surface tension (γ), with the wetting threshold

for CNTs being between 100 and 200 mNm-1[50]. Wetting with liquids with

a surface tension greater than the threshold is impossible.

42

It has been found that fluids with high surface tensions such as molten

metals are not successfully encapsulated inside CNTs except when forced to

do so by the application of very high pressures. Low surface tension solvents

have been found to effectively wet CNT and to be drawn into the interior of

the nanotube by capillary forces either via the open ends of the nanotubes

or through wall defects [50].

The nanotube wetting threshold is high enough that they are expected

to be wetted by water (γ ∼ 72 mNm-1) and most organic solvents (γ ∼72 mNm-1) [51]. The solubility of C60 and the relevant properties of some

solvents which have been used in nanotube filling experiments are shown

in Table 3.2. All of these solvents possess surface tensions low enough to

effectively wet carbon nanotubes and enter into the interior, however, super-

critical carbon dioxide (ScCO2) has no surface tension at all.

Solvent Solubility Critical diameter of Surface tension, γof C60 / gL-1 solvent molecule / nm at 20 o C / mNm-1

Toluene 2.800 0.78 28.52

Ethanol 0.001 0.44 22.39

Chloroform 0.160 0.33 27.32

n-Hexane 0.043 0.92 18.40

ScCO2 Unknown 0.28 0.00

Table 3.2: C60 Solubility, critical diameter and surface tensions of selected solvents[52–54].

It has been found that if a CNT is filled with a gas such as air - this will

oppose entry of liquid to the CNTs even if the liquid has a sufficiently low

surface tension to allow wetting of the tube [51].

43

To counteract this, it is necessary to provide the existing fluid with an

escape route such as the open ends of the nanotube or wall defects [51]. It

is also important that the solvent molecules once inside the nanotubes are

small enough to escape after the filling molecules have been encapsulated,

specifically they will have to fit through the defects in the nanotube side-

walls and around any pea-pod structures which have been formed. CO2 with

its linear form, has the smallest critical diameter of any of the molecules

listed in the table and therefore should have the greatest chance of escaping

from filled nanotubes.

The strength of the interaction between the solvent and the molecules

to be encapsulated can greatly affect the probability of encapsulation being

successful. If the solvent has a particular affinity for the molecule, as is the

case when C60 is solubilised in toluene, then the binding energy between

the solvent and solute can be so strong that encapsulation by the nanotube

becomes unfavourable [46]. In this instance the molecules will remain in

solution and be carried away from the nanotube. While the solubility of C60

in ScCO2 is not known it is expected to be low, but non-zero [55].

It is clear that with the above criteria in mind that the properties of

ScCO2 make it the ideal solvent for filling from solution experiments. Indeed,

it has been shown to be very effective in creating fullerene-CNT peapods

[15]. While the porphyrin and phthalocyanine molecules used in this study

are different from C60 geometrically, they share an aromatic structure. With

this in mind it is reasonable to assume that the above consideration should

apply to them also.

44

3.6 Molecular encapsulation by SWNTs us-

ing a ScCO2 medium

3.6.1 Supercritical fluids

A gas is characterised as having a low density, a low viscosity and a high

diffusity. In contrast a liquid has a relatively high density, a high viscosity and

a low diffusity. There is a third fluid phase which has properties intermediate

to those of liquids and gases; this is called a supercritical fluid (SCF). Table

3.3 shows the contrasting properties of the different fluids.

All supercritical fluids have a certain point in the phase space, charac-

terised by a critical temperature, Tc and a critical pressure, Pc at which they

make a transition from a normal fluid to a supercritical fluid; some examples

are shown in Table 3.4.

Phase Density / 103kg m-3 Viscosity / mPas Self diffusioncoefficient / 104m2s-1

Gas (0.6 - 2) × 10-3 (1 - 3) × 10-2 0.1 - 0.4

Supercritical fluid 0.2 - 0.5 (1 - 3) × -2 0.7 × 10-3

near to Tc

Liquid 0.6 - 1.6 0.2 - 3 (0.2 - 2) × 10-5

Table 3.3: The properties of fluids [56]

45

Fluid Critical temperature, Tc / o C Critical pressure, Pc / Bar

Helium-4 -268.0 2

Oxygen -118.6 50

Carbon dioxide 31.0 74

Propane 96.7 42

Table 3.4: Selected supercritical fluids and corresponding critical temperatures and pres-sures [57].

At this critical point (Tc, Pc) the boundary between the liquid and gas

phases ceases to exist and the liquid and gas densities become equal. Above

this point the fluid does not condense to form a liquid or evaporate to form

a gas, instead it has properties somewhere between the two [56]. Figure

3.7 shows the phase diagram of a single substance. Following the liquid-gas

coexistance curve from the triple point (T) to the critical point (C), both the

temperature and pressure of the fluid increases. The increase in temperature

results in thermal expansion of the liquid and as a result a decrease in density,

while the increase in the pressure results in an increase in the gas density.

Pressure

Temperature

Figure 3.7: Schematic diagram of the phases of a single substance (based upon[56]).

46

The fluid densities continue to converge until the critical point is reached.

At this point the densities equate and the liquid and gas phases become

indistinguishable [56]. With the disappearance of the barrier between the

gas and liquid phases, all inter-fluid surface tension is lost and the fluid

enters a supercritical phase.

The solvent power of a supercritical fluid is very sensitive to density fluc-

tuations, especially around the critical point. This means that relatively

small changes of pressure or temperature near to the critical point can be

used to tune the solvent power of the fluid. The low density and non-existent

surface tension of supercritical fluids make them an attractive prospect for

filling carbon nanotubes.

The critical temperatures and pressures of a selection of substances are

shown in Table 3.4. When deciding upon which substance to use in a super-

critical fluid experiment there are a number of practical factors to consider -

arguably the two most important are the temperatures and pressures of the

critical point. For example both oxygen and Helium 4 enter a supercritical

phase at relatively low pressures, but require cryogenic temperatures to come

near to the critical point. Substances exist which have critical points higher

in both temperature and pressure, however, they tend to be dangerous - for

example at its critical point propane is flammable and potentially explosive.

The substance most commonly used in supercritical fluid processes is

carbon dioxide. It has a critical pressure of 74 bar, well in range of modern

high pressure systems, and a critical temperature of 31.0 o C. These values are

such that CO2 can be maintained in the supercritical phase for an extended

period using relatively simple equipment with a low risk of adverse effects.

The low critical temperature of CO2 is especially useful when dealing with

temperature sensitive materials. Another useful characteristic of supercritical

CO2 is that it is chemically inert. This is important when using materials

sensitive to oxidation, such as the cobalt carbonyl molecule. Carbon dioxide

is also non-flammable, non-toxic, inexpensive and environmentally acceptable

[57].

47

3.7 ScCO2 induced encapsulation

Previous work [15] found that fullerenes of various sizes could be encapsulated

inside SWNTs using ScCO2 as a medium. In these studies the greatest

yield of encapsulated fullerenes was achieved when the ScCO2 was kept at

a constant temperature of 50o C and the pressure was cycled between a

pressure of 100 and 150 bars [15].

There are a number of factors that can determine the filling yield and

below is provided the rationale for the initial choice of the ScCO2 parameters

used for filling experiments.

The encapsulation process relies upon the solvent power of the ScCO2

which depends upon the density of the fluid. Figure 3.8 shows that the

density of a supercritical fluid is very sensitive to changes in temperature

and pressure near to the critical temperature, Tc = 31o C. The temperature

of 50o C used in the experiments of this study was such that the change in

density is fairly fast, but not as abrupt as near to Tc. For example, if T was

increased to a value that is significantly higher than Tc, for example over

100 o C, the change in density is more gradual, but it is also significanly less

dense - this would result in the fluid becoming more gas-like and adversely

affect the effectiveness of the fluid as a solvent, as well as adversely affect the

filling experiment. In contrast, if T is reduced below Tc the CO2 will leave

the SCF (supercritical fluid) state and re-enter the liquid phase, this would

be disasterous for a filling experiment. By changing the pressure of the fluid

while keeping the temperature fixed, it is possible to vary the density of the

fluid in a controlled way such that one of the isotherms is followed.

48

De

ns

ity

/ k

gm

-3

Pressure / Bar

0

200

400

600

800

1000

30 50 70 90 110 130 150 170

Figure 3.8: Schematic phase diagram of CO2 [56]. C is the critical point.

While the density of the fluid increases with pressure, the interaction

between the CO2 and the filling molecules does not vary in a straightforward

fashion. Figure 3.9 shows that at low densities (A), the interaction is weakly

attractive. As the density of the fluid is increased, a point (B) will be reached

at which the attractive interaction is maximised.

Figure 3.9: Dependency of the solute to solvent interaction as a function of supercriticalfluid CO2 density. The arrows indicate alternating cycle direction (adapted from [15]).

49

At this point the interaction is such that a thick shell of CO2 molecules

will form around the filling molecule (Figure 3.10 (a)). If the density is

increased further, the interaction will become increasingly less attractive and

eventually will become repulsive at (C). The repulsive effect will result in

a CO2-deficient region around the filling molecule (Figure 3.10 (b)). This

interaction is consistent with a van der Waals type potential, the density of

the fluid controlling the distance between the CO2 and filling molecules. By

ramping the pressure of the solution in this way it is possible to selectively

tune the solvent-solute interaction. When the pressure is cycled, the above

processes are repeated multiple times resulting in a higher filling yield.

RepulsiveAttractive

(a) (b)

CO2

filling molecule

Figure 3.10: Schematics showing (a) attractive, and (b) repulsive interaction betweenScCO2 solvent molecules and a solute (filling) molecule.

Further, information from previous work [15] shows that pressure-cycling

experiments produce much higher yields of encapsulated C60 molecules when

compared to static exposures of equal duration [15]. In a pressure cycling

experiment, the cycle shown in Figure 3.9 is repeated a number of times.

In the specific case of a solution of CO2 and C60 molecules, following

the curve from A to B, the inter-molecular interaction becomes increasingly

attractive until a CO2 density of≈ 440 - 620 kg/m3, i.e. at 85-95 bar for 40 oC

is reached at point B [55,58]. This attractive interaction causes solubilisation

of the filling molecules by the supercritical solvent (see part (a) of figure 3.10),

and allows them to be carried inside the nanotube. Inside the nanotubes,

the molecules will be irreversibly encapsulated.

50

Increasing the pressure further from B to C results in the interaction

becoming increasingly less attractive and eventually negative at C

≈ 750 kg/m3 (above 125 bar at 40 oC) is reached at point C [55, 58]. The

repulsive nature of the interaction at high pressure causes the solvent to

release the molecules (see part (b) of Figure 3.10). The solvent should now

decouple from the solute and leave the nanotubes. The range of pressures

used in the current investigation (100-150 bar) is consistent with this scenario.

Finally, there are some more general considerations regarding the filling

process. Due to the larger area of interaction between the encapsulated

molecules and the nanotube inner walls, their interaction energy can be very

high and can result in irreversible encapsulation inside the tubes.

In contrast, the interaction energy between the CO2 and the nanotube is

much less, and this makes their trapping much less likely: the small diameter

of the CO2 molecules, of only 2.8 A, allows them to diffuse around the filling

molecules, and leave the nanotube through the open nanotube ends or side

wall defects in the tube.

51

Chapter 4

Production of hybrids of

nanotubes and organo-metallic

molecular systems

This chapter describes the supercritical CO2 instrumentation and processes

developed and implemented in order to encapsulate the chosen organo-metallic

systems. This resulted in two variants, a primary one, suitable for air-

stable compounds, and another designed for dealing with air-sensitive sys-

tems. Preparation procedures applied to the carbon nanotubes prior to the

supercritical fluid encapsulation processes, and associated characterisation

are also described. Finally, a protocol for the exohedral functionalization of

carbon nanotubes is also given.

4.1 Synthesis of carbon nanotubes

The three methods of carbon nanotube synthesis most commonly employed

are laser vaporization, arc discharge and carbon vapour deposition (CVD).

Each method is quite different and tends to produce SWNTs with different

average diameters with different diameter distribution widths and different

levels of impurities.

52

For the experiments in this study a synthesis method which produces a

broader diameter distribution is desirable. This is because while the the-

ory discussed in the previous chapter provides an estimate for the optimum

filling diameters for each molecule, they are not exact. A broader diameter-

distribution should increase the probability of the optimum diameter for each

molecule being present in the nanotube samples and provide the opportunity

of forming new supramolecular architectures.

4.1.1 Laser vaporization

In the laser vaporization method graphite containing a small amount of

embedded transition metal catalyst particles is vaporized using a pulsed laser

and condensed into SWNTs. This method tends to produce relatively small

amounts of SWNTs and the diameter distribution curve tends to be very

narrow [3]. This method would be very useful in experiments where a specific

nanotube diameter is desired.

4.1.2 Arc-discharge

One of the most commonly used methods of nanotube growth, in the arc-

discharge - method a voltage is applied across two graphite rods separated

by a ≈ 1 mm gap. In order to grow single-walled nanotubes, it is necessary

to embed catalyst particles in the graphite rod which acts as the cathode -

it is upon these particles that the nanotubes grow. The catalyst particles

are usually nanoparticles of transition metals such as Co, Ni or Fe [3]. In

addition to nanotube deposition, fullerenes and amorphous carbon are also

deposited onto the cathode at the same time [3]. The arc discharge method

is a high temperature process which requires temperature of ≈ 3000 o C [3].

The exact properties of the nanotubes that are grown by this method will

depend upon the conditions used; however, in general, the average diameter

of the nanotubes and the width of the distribution curve tends to be small.

The arc-SWNTs used in the experiments of this study were commercially

grown by NanoledgeTM and they have an average diameter of 1 nm.

53

These nanotubes were chosen because their average diameter matches

well with the expected optimum diameter (0.8 nm) of the cobalt carbonyl

molecules as discussed in the previous chapter.

An example of a nanotube diameter distribution obtained using the arc-

discharge method is shown in Figure 4.1 [59]. This distribution shows that

although more nanotubes were produced with a diameter of ≈ 1.05 nm, a

significant fraction were produced which have diameters close to the peak di-

ameter. The diameter distribution is skewed to the left of the peak diameter,

favouring narrower nanotubes. This tendency would increase the number of

nanotubes with the optimum encapsulation diameter for the cobalt carbonyl

molecule and hence could result in an increase in filling yield.

Figure 4.1: A histogram showing the number of nanotubes produced as a function ofnanotube diameter for an arc-discharge method using an Fe catalyst [59].

54

4.1.3 Carbon vapor deposition

The carbon vapor deposition process uses a lower temperature than the arc

synthesis method. In this method a gas of carbonaceous material is passed

over catalyst nanoparticles, usually of transition metals which are kept at a

temperature of ≈ 1000 oC in a reaction tube [3]. The CVD method tends

to produce SWNTs with larger diameters and broader diameter-distributions

than the arc method. Due to the lower temperature, the nanotubes also tend

to have a larger density of defects in their structure [3]. In addition to the

usual amorphous carbon and catalyst particle impurities, the CVD process

tends to result in a greater number of double and multi-walled nanotubes

being produced. The CVD nanotubes used in this study were sourced from

NanocylTM. They have an average diameter of 2.0 nm [60] - this is consis-

tent with the approximate optimum filling diameter of the porphyrin and

phthalocyanine molecules.

Figure 4.2 shows the diameter distribution for SWNTs grown by the

CVD method using Fe2O3 nanoparticles as a catalyst [61] - used here for

qualitative comparison with the Nanocyl nanotubes. Comparing Figure 4.2

with Figure 4.1, it can be seen that the diameter distribution obtained by the

CVD method is broader than that produced by the arc-discharge method.

The greater breadth of CVD distributions means that there should be a

significant number of nanotubes with diameters slightly greater and slightly

lesser the peak diameter - this should result in there being an ample supply

of nanotubes with diameters near to the optimum filling diameters of the

porphyrin (2.3 nm) and phthalocyanine (1.8 nm) molecules used in this study.

55

Figure 4.2: A histogram showing the number of nanotubes produced as a function ofnanotube diameter for an CVD method using an Fe2O3 catalyst [61].

4.2 Purification of carbon nanotubes

4.2.1 Impurities

As-grown carbon nanotubes typically possess a number of impurities, which

in the case of CVD-grown nanotubes, can amount to as much as 30 % of

the batch material [60]. These impurities are composed of particles of amor-

phous carbon, residual metallic catalyst particles and carbon onions. A car-

bon onion is a metal catalyst particle around which a shell of amorphous

carbon has formed. A number of purification methods have been developed

to remove these impurities - these methods are now described:

(i) 4.2.2 Thermal oxidation

Thermal annealing in an oxidising atmosphere has been shown to be

an effective way to remove amorphous carbon from carbon nanotube

batches [62]. In this process the nanotube powder is heated at a tem-

perature in excess of 350o C in air. This process exploits the weaker

disordered bonding in the amorphous carbon structures relative to the

strong sp2 lattice bonding in the nanotubes.

56

The combination of increased reactivity of the weaker bonds at high

temperature and the presence of oxygen in the air results in the amor-

phous carbon being oxidised and removed as either carbon dioxide or

carbon monoxide. This process is effective in removing both amor-

phous carbon particles and the shells of carbon around the carbon

onions (where reactivity is increased due to the high curvature). The

main drawback of this process is that it can also damage the nanotubes,

especially if the

nanotubes possess defects [63].

(ii) 4.2.3 Hydrogen peroxide-based oxidation

Another process which has been shown to be effective in removing

amorphous carbon from as-grown nanotubes is oxidation by hydrogen

peroxide (H2O2) in solution [63]. In this process the H2O2 oxidises the

amorphous carbon producing either carbon dioxide or carbon monox-

ide. This method is less aggressive than thermal annealing and hence

will result in less damage to the nanotubes.

(iii) 4.2.4 Acid reflux

A simple and effective method to remove residual catalyst particles

is to reflux the powder in concentrated hydrochloric acid (HCl) [64].

The HCl dissolves the metal catalyst particles, but does not attack the

nanotubes.

4.2.5 Purification procedures adopted

Using a combination of carbon oxidation and acid reflux processes it is possi-

ble to remove a great deal of the impurities from as-grown carbon nanotube

batches.

Two purification procedures were adopted for application to the nan-

otubes batches used in the experiments of this study.

57

The first was more aggressive, and employed a combination of thermal

annealing and acid reflux techniques. Use of this procedure resulted in a

substantial reduction of the overall mass of purified nanotubes.

The sequence of steps is as follows:

Purification procedure 1

(i) As-grown nanotubes were heated in a solution of 70% concentrated HCl

for one hour at a temperature of 105o C in order to remove any exposed

catalyst particles.

(ii) The suspension was then filtered with deionized (D.I.) water under

vacuum pumping to dilute and remove the acid and left to dry at 90o

C over-night.

(iii) The resulting powder was heated in air for 45 minutes at 450o C to

oxidize the amorphous carbon and onion cages.

(iv) The dry powder was heated in a solution of 70 % concentrated HCl for

one hour at a temperature of 105o C in order to remove the metallic

cores of the carbon onions expsoed by the oxidation process.

(v) The suspension was then filtered with D.I. water under vacuum pump-

ing to dilute and remove the acid and left to dry at 90o C over-night.

Purification procedure 2

The second purification procedure used a combination of the hydrogen per-

oxide oxidation and acid reflux processes. This procedure results in a milder

and more selective purification, and a larger quantity of purified nanotubes

is obtained.

The sequence of steps is as follows:

(i) As-grown nanotubes were heated in a solution of 70% concentrated HCl

for one hour at a temperature of 105o C.

58

(ii) The suspension was filtered with D.I. water and enthanol mixture under

vacuum pumping and left to dry over-night at 90o C.

(iii) The resulting powder was gently sonicated using a low-powered ultra-

sonic bath in 30 % concentrated H2O2 and then left to stir for one

month.

(iv) The suspension was filtered with D.I. water (to dilute and remove any

remaining H2O2) and enthanol under vacuum pumping and left to dry

over-night at 90o C.

(v) The dry powder was heated in a solution of 70 % concentrated HCl for

one hour at a temperature of 105o C in order to remove the metallic

cores of the carbon onions exposed by the oxidation process.

(vi) Step (ii) was then repeated.

4.3 Characterisation of purified nanotubes

Before filling experiments using the commercial arc (Nanoledge) and CVD

(Nanocyl) nanotubes were attempted, the quality of the nanotubes was in-

vestigated using transmission electron microscopy (TEM). The TEM images

obtained, which are discussed below, made it clear that further purification

of the nanotube material was necessary - therefore, both types of nanotube

were subjected to purification to remove impurities introduced during the

production process.

4.3.1 Characterisation of Arc SWNTs

The Nanoledge arc SWNTs were subjected to purification procedure 1 de-

scribed above. From the TEM images (Figure 4.3) of the nanotube mate-

rial before and after purification it can be seen that both the amount of

amorphous carbon and the number of catalyst particles are greatly reduced

post-purification.

59

(a) (b)Amorphous carbon

Catalyst particle

Nanotube bundle

keV

Fe

Si

Cu

C

Figure 4.3: Transmission electron microscope (TEM) images of (a) as-grown commer-cial arc (Nanoledge) SWNTs and (b) after purification using procedure 1. The energy-dispersive x-ray (EDX) spectrum inset into part (b) shows that the arc-discharge nanotubesample still has a significant number of Fe catalyst particles even after purification. The Cusignature originates from the copper TEM grids on which the nanotubes were deposited.The small Si peak is another impurity left over from the growth process.

Amorphous carbon can be identified as irregular lumps of material with

a contrast similar to that of the nanotubes. The metallic catalyst particles

appear as very dark spots - due to their relatively high atomic number they

tend to scatter the electrons of the beam and hence very few electrons are

transmitted. The level of magnification shown in these images is too low to

see individual SWNTs, and only bundles of nanotubes are visible.

4.3.2 Characterisation of CVD SWNTs

The commercial CVD nanotubes used in this study were purified using both

procedures 1 and 2. Figure 4.4 shows the nanotubes before and after purifi-

cation using method 1.

60

(a) (b)

keV

C

keV

Fe

C

Figure 4.4: (a) as-grown commercial CVD (Nanocyl) SWNTs and (b) after purificationusing procedure 1. The EDX spectrum inset into parts (a) and (b) shows that the CVDnanotube sample still has Fe catalyst particles present before purification and a negligiblenumber after purification.

After purification using this method there is a clear reduction in both the

amount of amorphous carbon and the number of metallic catalyst particles

present in the material. Comparing figures 4.4 and 4.3 one can see that

the CVD nanotube sample has less metallic particles present than the arc

nanotube sample purified with the same protocol. This is due to the presence

of a large amount of onions with metallic cores which are produced during

the arc-discharge synthesis process, which are difficult to remove.

Figure 4.5 shows the nanotubes before and after purification using method

2.

61

(a) (b)(b)

Figure 4.5: (a) as-grown comercial CVD (Nanocyl) SWNTs and (b) after purificationusing procedure 2

This method also resulted in a significant decrease in the amount of im-

purities in the material, however, perhaps not as much as achieved using the

high temperature annealing process used in procedure 1.

Although purification procedure 1 appears to be more efficient than pro-

cedure 2, it was decided that it would be better to use procedure 2 to produce

the purified nanotubes to be used for the Raman spectroscopy experiments of

this study. It was thought that the hydrogen peroxide-based process would

result in fewer defects in the structure of the nanotube and hence create less

disruption to the nanotube spectra.

4.4 Nanotube end-opening

When formed, carbon nanotubes have hemi-spherical fullerenes capping their

ends. These present a significant barrier to filling molecules and therefore

need to be removed before a filling experiment is conducted. A method

known to remove these end caps is to heat the nanotubes in air [65].

62

This process exploits the weaker C-C bonds in the in the end-caps of

the nanotubes - the greater curvature and hence greater reactivity results in

the carbon atoms in the caps being oxidised by the oxygen in the air and

removed. The curvature of the walls of the nanotubes is less than that of

the end-caps and hence the walls are stronger and less reactive - this makes

them less susceptible to the oxidation process. The main draw-back of this

process is the high temperature necessary to remove the end-caps can also

damage the walls of the nanotubes. However, this process is necessary if a

reasonable filling yield is to be obtained. Before each of the filling experiments

conducted in this study the nanotubes were heated in air for 45 minutes at

a temperature of 450 o C, to remove the end caps. The nanotubes to be

covered with molecules were not subjected to this process because filling was

not the goal.

4.5 Supercritical fluid molecular filling

experiments

4.5.1 Equipment set-up

Governed by the specific properties of the molecular systems to be filled, two

configurations of the rig have been used. Configuration A (Figure 4.6) is a

primary one, compatible with the use of molecules that are not air-sensitive.

Alternatively, for molecules that oxidize on exposure to air, configuration B

(Figure 4.8) was required. This involved further development of configuration

A, so that molecules were kept in a dry solvent (i.e avoiding air exposure)

over the whole duration of the filling experiment.

The purpose of the experimental rig shown is to provide and maintain

a stable supercritical CO2 fluid environment with specific user-set pressure

and temperature conditions. The various parts of the rig are indicated in the

figure and their roles are explained below.

63

4.5.2 Configuration A

Figure 4.6: Schematic diagram of the supercritical CO2 rig in its basic configuration, A.

The liquid CO2 is supplied by a cylinder which is at room temperature and

pressurised to approximately 63 bars. Before entering the CO2 pump, the

liquid CO2 is cooled to 0 o C, it is then compressed by a liquid-CO2 pump. In

order to access the supercritical fluid state, the CO2 needs to be compressed

to a pressure greater than the critical pressure of 74 bars. After compression,

the liquid CO2 is first heated above the critical temperature of 31 o C by a

dedicated CO2 heater and then again in the autoclave. The autoclave is a

hollow stainless steel cylinder into which the nanotube samples to be filled are

inserted. A thermocouple and dedicated heater maintain the sample region

at a constant, chosen temperature. A back-pressure regulator maintains

a constant system pressure at a user-set value by opening and closing an

incorporated needle valve. At the end of the experiment, the waste CO2

is exhausted through the back pressure regulator to the outside. During

the molecular filling experiments conducted using this apparatus, the system

pressure was taken through a number of repeated cycles as explained in

chapter 3.

The majority of the equipment from which the rig is constructed is com-

puter controlled - this is a very useful feature as it makes lengthly sequences

of exact pressure cycles possible. An example of such a pressure cycle is

shown in Figure 4.7.

64

0

20

40

60

80

100

120

140

18:00 18:20 18:40 19:4019:2019:00 20:00 20:20 20:40 21:00 21:20 21:40

Figure 4.7: Example pressure cycling experiment

Typically, in these experiments a cycle started at a pressure of 100 bars,

ramped up to 150 bars and then returned back to 100 bars. The period

of time and number of the cycles is different depending on the type of ex-

periment being run. The choice of experimental conditions will be justified

in the next section, and depends on the physical processes that govern the

nanotube filling.

4.5.3 Configuration B

Figure 4.8: Schematic diagram of the rig in the configuration B, for filling from solution.

65

The filling molecules used in the second type of experiments are extremely

sensitive to oxidation. While the experiment is in progress, the solution in

which the molecules are dissolved (shown in the photograph in Figure 4.9 as

a dark yellow liquid) is stored in an air tight experiment flask. The flask is

connected to a supply of argon, this maintains a constant inert atmosphere

throughout the experiment. To further reduce the risk of contamination, it

was necessary to connect a bubbler to the gas line. This prevents a sudden

vacuum from drawing air into the flask and contaminating the solution. The

other major additions to the system are a co-solvent pump and a solvent

mixer. The co-solvent pump is used to draw the molecule solution from the

flask and to add it to the CO2 pipeline at the mixer. The co-solvent pump is

computer controlled and capable of automation. The mixer simply mixes the

relatively low pressure solution with the high pressure CO2. A photograph

of the whole set-up is shown in Figure 4.9.

Figure 4.9: Filling from solution experiment in progress with the equipment in configu-ration B.

66

4.6 Sample production

Three different types of nanotube/ molecule hybrids were produced: (a) nan-

otubes endohedrally supercritically-filled from a mixed nanotube/ molecule

powder, (b) nanotubes endohedrally supercritcally-filled from a molecular

solution and (c) nanotubes exohedrally functionalised on the outer surface

with molecules. Nickel tetra phenyl prophyrin (NiTPP), Nickel Phthalo-

cyanine (NiPc) and Aluminium Phthalocyanine Chloride (ClAlPc) molecule

powders were purchased from Sigma Aldrich.

4.6.1 (a) ScCO2 filling of nanotubes from powder

This type of experiment involved supercritical filling of CVD SWNTs with

the NiTPP, ClAlPc and NiPc organo-metallic molecules. Given that these

molecules are not air sensitive, a powder mixture approach was used, with the

nanotubes and molecules being mixed prior to being exposed to the ScCO2.

Configuration A of the supercritical CO2 rig was used for the experimental

set-up. In this case, the ScCO2 helps the molecules to diffuse inside the

nanotube mat and reach the open ends of the nanotubes, from where they are

further carried inside the hollow cavity of the nanotube by the supercritical

fluid.

Several stages of material preparation were necessary before running the

experiment:

(i) CVD SWNTs were purified using either purification method 1 or 2 de-

pending upon whether the sample was to be used for high resolution

transmission electron microscopy or resonant Raman studies respec-

tively.

(ii) The purified nanotubes were end-opened using the procedure discussed

earlier.

(iii) The molecules to be used in the filling experiment were dissolved in

the solvent appropriate to each individual molecule (see Table 4.1) to

produce a saturated solution.

67

Both the NiTPP and ClAlPc dissolved easily in the solvents suggested

by literature [39,66]. However, the NiPc molecule did not dissolve well

in any of the solvents tried; it was found to partially dissolve in ethanol.

(iv) The purified nanotubes were added to the molecule solutions and to-

gether they were mixed by mild sonication and mechanical stirring.

(v) This solution was then drip condensed onto a piece of silicon substrate

held on a heated plate at a temperature above the evaporation temper-

ature of the solvent, forming a nanotube/ molecule mat (Figure 4.10

(a)).

Molecule Solvent

NiTPP Chlorofrom

ClAlPc Ethanol

NiPc Ethanol

Table 4.1: Solvents used for solubilising the molecules [39,66].

When the mat had dried, the silicon substrate was then attached to a

sample holder for insertion into the system autoclave. The sample holder

rests on the lower frit of the autoclave as in Figure 4.10 (b). The frits are

filters with micro-meter sized pores, the lower frit prevents any contamination

from the pumps or pipeline from contaminating the system, while the upper

frit prevents clumps of the sample from escaping from the autoclave.

Clamps

Silicon substrate

Sample holderNanotube / molecule mat

Caps Frits

Sample

holder

Autoclave

(a) (b)

Figure 4.10: (a) Schematic diagram of the sample holder prior to autoclave insertion.(b) cross-section view of the autoclave with the sample inserted.

68

Pressure cycles between 100 and 150 bars, similar to those shown in Figure

4.7, were run. A typical experiment lasted for a number of days with a

number of pressure cycles being carried out; a summary of the experiments

conducted is given in Table 4.2. The experiments were run with a fixed

sample temperature of 50o C.

4.6.2 Summary of samples produced

Five samples were produced, the two samples destined for HR-TEM study

were purified using procedure 1, had shorter filling experiments with fewer

pressure cycles and used only the NiTPP molecule. The remaining three

samples were purified using procedure 2, had longer filling experiments with a

greater number of cycles, and all three of the non-air sensitive molecules were

used. These samples were studied using Raman spectroscopy. A summary

of the samples produced is given in Table 4.2.

Molecule Number of cycles Cycle duration / hours Future experiments

NiTPP 3 12.5 HRTEM

NiTPP 3 30 HRTEM

NiTPP 10 24 Raman

ClAlPc 9 24 Raman

NiPc 8 24 Raman

Table 4.2: Summary of samples produced by the powder filling experiments.

4.6.3 (b) ScCO2 filling of nanotubes from a molecular

solution

For filling of SWCNTs with Cobalt Carbonyl molecules it was necessary to

develop a method which avoided contact of the molecules with oxygen, this

is because the molecules readily oxidise in air.

69

For this reason it was decided to use a filling-from-solution method, with

the set-up in configuration B. In contrast to the powder filling experiments,

the filling molecules in solution are introduced directly to the ScCO2. It was

thought that this might have resulted in a greater solubilisation of the cobalt

carbonyl molecules and hence might allow the molecules to diffuse through

the nanotube mat more easily. Cobalt carbonyl molecules were purchased

from Sigma Aldrich.

The nanotubes of this experiment were prepared using the following pro-

cedure:

(i) Arc nanotubes material was subjected to purification procedure 1.

(ii) The nanotubes were then end opened using the procedure described

earlier.

(iii) Hexane was used to make a suspension of the nanotubes, and this

suspension was drip condensed to form a mat on a silicon substrate

which was attached to the sample holder.

The main differences between this method and that of the method used

in the powder filling experiments is that the molecules were not mixed with

the nanotubes directly, and the use of smaller diameter arc-nanotubes.

The creation of the cobalt carbonyl solution using dry hexane (dry hexane

being oxygen free) required a completely new method of sample preparation.

The oxygen sensitive nature of the filling molecules meant that all the pro-

cessing stages had to be done in an inert atmosphere. The inert atmospheres

for the weighing of the molecules and the subsequent mixing of the molecule

solution were provided by the use of a glove box and Schlenk line under a

nitrogen atmosphere respectively. The resulting solution was kept under an

argon atmosphere for the entire filling experiment, as shown in Figure 4.9.

The cobalt carbonyl/ hexane solution was slowly added to the supercritical

CO2 during the filling experiment by the co-solvent pump via the mixer.

The type of co-solvent pump used for these experiments is designed to pump

solvents such as hexane.

70

The pressure of the CO2 was cycled between 100 to 150 bars as with the

other powder experiments, however, due to the limited volume of cobalt car-

bonyl / hexane solution contained in the experimental flask the experiments

were limited to cycles of only 50 minutes in duration.

Unfortunately, these experiments were unsuccessful and no useable sam-

ples were produced. This resulted from the cobalt carbonyl/ hexane solution

causing the valves in the co-solvent pump to block and hence causing the

pumping to stop. It was necessary to dismantle part of the apparatus and

send the pump away for costly repairs. In an attempt to overcome this

problem, a number of these experiments were conducted, with both dilute

and very dilute solutions; however, the result was the same each time. It

was decided that in order to save time and avoid costly repairs, that it was

best to concentrate on the experiments using the air-stable molecules and to

abandon air-sensitive cobalt carbonyl molecules as filling materials.

4.6.4 (c) Exohedral functionalisation of SWNTs

These experiments produced CVD SWNTs with their exterior surfaces cov-

ered with each of the air-stable molecules. The procedure used to create

these samples is as follows:

(i) CVD SWNTs were purified using purification method 2.

(ii) The molecules to be used to produce the nanotube hybrid samples

were dissolved in the solvent appropriate to each individual solvent

(see Table 4.1) to produce a saturated solution.

(iii) The purified nanotubes were added to the molecule solutions and to-

gether they were mixed by mild sonication and stirred mechanically for

several days.

(iv) This solution was then drip condensed onto a piece of silicon substrate

held on a heated plate at a temperature above the evaporation tem-

perature of the solvent, forming a nanotube/ molecule mat.

71

A method similar to the one described has been found to result in nan-

otubes with a good coverage of molecules [42].

The key differences between the samples production methods for covered

samples are that the samples were not end-opened prior to being covered

and the covered SWNTs were not exposed to ScCO2. The exclusion of these

two steps should ensure that the number of molecules attached to the outer

surface of the nanotubes should far exceed those encapsulated inside which

may have entered through defects in the walls of the nanotubes. The samples

produced by this method were investigated by resonant Raman spectroscopy.

It is important to note here that the nanotubes used to create these

samples were from the same batch as those used to make the ScCO2 filled

from powder samples, matched by molecule. This means that it is reasonable

to compare the Raman spectra of SWNTs filled and covered with the same

molecular system. However, the nanotube batches used to form the hybrid

samples were dissimilar between molecular types and therefore cannot be

compared reliably. This is because a Raman spectrum acquired from SWNTs

from batches which have undergone a greater amount of purification may be

slightly different to a spectrum acquired from to an earlier batch which is

less purified.

4.7 Removal of extraneous molecular

material

After each nanotube filling or covering experiment, the samples were washed

using the appropriate solvent for each individual molecule Table 4.1. This

process was necessary to remove excess (i.e. unattached) molecules in the

sample. Excess molecules can cause the resonant Raman spectra of nan-

otube/ molecule hybrid samples to be dominated by the spectrum of the

molecules. The washing was effected by filtering the sample with the ap-

propriate solvent (i.e. the solvent which solubilised the respective molecule)

under vacuum pumping until the filtered solvent became colourless. The

resulting powder was then washed with water and dried over-night at 90o C.

72

Chapter 5

HRTEM investigations of the

internal structure of hybrids of

nanotubes and organo-metallic

molecular systems

We demonstrate through High Resolution Transmission Electron Microscopy

(HRTEM) the successful encapsulation of large, planar organo-metallic molecules

(larger than in previous works) by SWNTs using a supercritical CO2 based

method. Due to their size (≈2 nm across), such molecular systems are ex-

pected to be difficult to encapsulate using more standard, thermal diffusion-

based procedures. Encapsulation was obtained in nanotubes with a wide range

of diameters, including diameters less than an optimum value resulting from

geometrical considerations. Unlike in other works, HRTEM revealed row-like

ordering of the organo-metallic molecules in nanotubes with diameters close

to the optimum value. Confinement by the nanotube template appears to play

a role as templates of larger diameters did not induce ordering.

73

5.1 High Resolution Transmission Electron

Microscopy

High resolution transmission electron microscopy (HRTEM) is a very useful

technique for studying carbon nanotubes - it can be used to identify the

presence of nanotubes in a sample, perform measurements of the diameters

of nanotubes in a sample to provide statistics for diameter distributions and

to identify structural defects [23].

Transmission electron microscopy is a direct imaging technique [23] which

allows for the interior of the nanotubes to be probed. This makes it ideally

suited to characterising the internal structure of carbon nanotube hybrids.

Figure 5.1 shows a schematic drawing of the most important components

of a TEM.

Electron gun

(source)

Condenser lens

Specimen

Objective lens

Projector lens

Imaging screen

Figure 5.1: Schematic diagram showing the primary components of a TEM, includingthe most important magnetic lenses, specimen and imaging screen (based upon [67]).

74

A TEM works by exploiting the particle wave duality of electrons, which

is described by the de Broglie relationship:

λe =h

mev, (5.1)

where λe is the effective wavelength and v is speed of the electrons in the

TEM beam, me is the mass the electron, and h is Planck’s constant [67].

With sufficiently high electron velocities it is possible to obtain effective

wavelengths of less than an angstrom, and hence to resolve structures much

smaller than possible with an optical microscope.

Electrons are accelerated in the electron gun of the TEM by a voltage

V . This is the source of the radiation which will illuminate the sample to

be investigated. An electron accelerated by a voltage V will gain a kinetic

energy described by 12mev

2 = eV , where e is the charge of the electron.

From which: v =√

2eVme

. Substituting this equation for v in equation 5.1

gives: λe = h√2meeV

. Putting in the values for the constants, the following

expression is obtained: [67]:

λe =12.3√V

A. (5.2)

It is clear that the larger V becomes the smaller the effective wavelength of

the electrons becomes. For example for an acceleration voltage of 100 kV,

λe = 0.039 A. If the resolution of a TEM depended entirely upon the effective

wavelength of the electrons being accelerated, it could achieve sub-atomic

resolution. However, the main limit to the resolution of a TEM is the quality

of the magnetic lenses which guide and focus the electron beam.

The condenser lens regulates the convergence of the illuminating beam

on the specimen. The objective lens focuses the electron beam which has

passed through the specimen and provides the first magnification of the image

produced. The projection lens magnifies a portion of the magnified image

further to form the final image.

75

The lens which has the greatest effect on the resolution of the image

formed is the objective lens; it is the most critical component of a TEM [67].

There are a number of intrinsic aberrations to electron optics which cannot

be corrected, but their effects can be minimized. The most important is

known as spherical aberration [67] - it is produced by the geometry of the

lens field. It occurs along the axis of the beam and results from the lens

further away from the centre of the beam having a greater refractive power

and hence shorter focal length. In a modern HRTEM system with a suitable

sample it is possible to obtain minimum resolutions in the order of a few

angstroms.

Images are obtained from a TEM by an imaging system which converts

the electron radiation into visible light such as a fluorescent viewing screen

[67].

Details of the images produced are due to variations in specimen contrast

[67]. When the electrons of the beam transit through a specimen, a fraction

will be scattered by the material present. If scattered by a large enough

angle, the electrons are lost from the beam and a corresponding loss of beam

intensity results. The amount of scattering which occurs at a given point

in the specimen is dependent upon the physical density and thickness of the

material there. In the images shown in this chapter, bright regions indicate

high electron transport and low scattering, while dark regions indicate a high

level of electron scattering and a low level of transmission.

In the TEM study of carbon nanotubes the factor most likely to result

in specimen damage is exposure of the nanotubes to the electron beam.

This is especially true for nanotubes studied under HRTEM conditions using

acceleration voltages in excess of 100 keV.

Damage to the sample occurs through inelastic scattering between the

electrons of the beam and the material of the specimen. The amount of

damage done to the sample depends most of all upon the beam current

and the exposure time. The minimum beam intensity required to form a

usable image limits how far the beam current can be reduced, therefore when

imaging samples it is wise to minimize the duration for which the sample is

exposed to the beam.

76

Irradiation of carbon nanotubes by an electron beam results in structural

defects and eventually destruction of the nanotube.

5.2 Equipment and experimental methods

SWNTs filled with NiTPP molecules were sonicated in chloroform using an

ultra-sonic probe to form a well dispersed and de-bundled solution. Drips

of this solution were deposited onto lacey carbon film TEM grids. Lacey

carbon film grids are formed from very thin films of carbon which possess

irregular holes. The best contrast of the interior of the filled nanotubes is

obtained from individually separated nanotubes which are suspended across

these holes.

HRTEM was performed by Dr. Adelina Ilie using two microscopes at the

University of Oxford, a JEOL 3000F field-emssion gun (FEG) microscope

and a 4000HR with a LaB6 source, both operated at 100 keV. The 3000F

microscope had a spherical aberration coefficient (Cs) of 0.57 mm giving

it a point resolution of ≈ 0.225 nm at 100 keV, while the 4000HR had a

substantially less good resolution due to higher energy spread and beam

divergence (see Table 5.1). Nevertheless, the resolution of the 4000HR was

sufficient to allow one to observe the contour of the encapsulated molecules

(see section 5.4). The majority of the images in this study were taken with

the 4000HR microscope.

77

Microscope Spherical Defocus Energy Beamaberration spread / nm spread / eV divergencecoefficient / mm / mrad

4000HR 0.9 10 1 1(LaB6 source)

3000F 0.6 4 0.1 0.15(FEG)

Table 5.1: Microscope parameters for the two HRTEMs used in this study.

The 100 keV energy of the electrons is slightly above the knock-on thresh-

old of ≈86 keV for carbon atoms in pure carbon nanostructures [68,69], such

as empty SWNTs, and is considered as appropriate for routine images of such

nanostructures [68]. In order to reduce the rate of knock-on displacement of

carbon atoms [69], the irradiation dose was kept low. Calculations based

upon Ref. [69] show that at 100 keV, under beam current densities of 0.4

to 2 A/cm2 (as used here), the displacement rate of carbon atoms, p, ranges

between 3.6 × 10−5 to 1.8 × 10−4s−1; this is equivalent to saying that each

carbon atom under the beam has been displaced once in 1/p ranging from

28000 s (for the lowest dose), to 5600 s (for the highest dose) [70]. These val-

ues compare well with our observation times, which were kept low, typically,

to only several minutes.

To compare with the experimental images, HRTEM images were sim-

ulated. For this, structural models of carbon nanotubes with encapsulated

organo-metallic molecules were generated with Crystal Maker [71], while their

image simulations were performed with SimulaTEM [72]. Sets of focal series

were produced with the molecules in different conformations.

Generating focal series implied that the focal length was varied from under

focused to over focused. This was done in order to minimise the effect of the

spherical aberration Cs. At the so called Scherzer focus zs given by:

78

zs = −c√Csλ (5.3)

with c a constant between 1.0 and 1.2, the image of a weak phase (thin)

object, such as a carbon nanotube is intuitively interpretable [73].

5.3 HRTEM of related systems

Prior work on organometallic systems close in size to those used in this study

will now be discussed. For all encapsulated systems, when determining the

optimum filling diameter for a particular molecule it is necessary to consider

both the geometry of the molecule and the interaction between the van der

Waals surface of the molecule and that of the interior of the nanotube in

which they are to be encapsulated. In general the optimum filling diameter

of the nanotube is given by the equation 3.3.

In a study by Schulte et al [36], roughly square CoPc molecules with

a corresponding van der Waals surface of ≈ 1.1nm2 in area [74] and 1.5

nm across the diagonal were encapsulated inside carbon nanotubes [36] (see

Figure 5.2 (a)). With the dimensions quoted, the CoPc molecule would have

doptimum of ≈ 1.4 nm for side-on filling (Figure 5.2 (c)) and an doptimum≈1.8 nm for face-on filling (Figure 5.2 (d)). CoPc molecules were found to

be encapsulated with a high yield inside of nanotubes of 2.2 and 2.6 nm in

diameter and to a lesser extent inside of nanotubes with d = 1.5 nm [36]. The

higher filling yield observed in the larger diameter nanotubes is attributed to

greater freedom of entry for the molecules into the interior of the nanotubes

[36]. The authors also deduced that in nanotubes with d < 1.8 nm the

molecule will only fit inside the molecule if inclined at an angle. We note

that this might be due to the rather rigid structure of the phthalocyanines;

the six-member ring appendages are linked through two C-C bonds to the

phthalocyanine body which does not allow significant distortion outside of

the plane of the molecule.

79

The authors did not observe any significant ordering of the CoPc encapsu-

lates in any of the nanotube templates, from narrow to wide diameter, based

on HRTEM evidence (see Figure 5.2 (a)), though some ordering was inferred

from near-edge X-ray absorption fine structure (NEXAFS) investigations.

(a) (b)

a

b

c

0.15 nm

1.5 nm 1.8 nm

0.15 nm

1.4 nm1.1 nm

(c) (d)

Figure 5.2: HRTEM images of carbon nanotubes encapsulated with CoPc molecules [36](a) and H8Si8O12 [34] (b). Schematic diagrams of the molecules [34, 36], are shown asinserts. (c) and (d) show side-on and face-on contours of the CoPc molecule inside of anoptimum diameter SWNTs respectively.

In a similar study by Wang et al [34] both SWNT and MWNTs were filled

with roughly cubic H8Si8O12 molecules (see Figure 5.2 (b)). The van der

Waals diameter of the molecule was calculated to be 0.9 nm with a resulting

optimum diameter of 1.2 nm across the diagonal of one side of the cube [34].

They found that SWNTs of greater than 1.2 nm were filled to a high degree

with H8Si8O12 molecules. The lack of aromatic appendages meant that no

self-assembly inside of the nanotubes templates occurred for these molecules.

80

Both systems shown in Figure 5.2 are being used for comparison with our

own experimental images.

5.4 Structural characterisation of hybrids of

SWNTs and endohedral NiTPP

The NiTPP molecule is a roughly square molecule, its van der Waals surface

measures 2.0 nm across the diagonal of the molecule and 1.4 nm along the

side - the molecule is shown schematically in Figure 5.3 (a).

With these considerations in mind, there will be two optimum diameters

for the NiTPP molecule, one measured using the dimension of the side of the

molecule and another using the diagonal. A molecule that might not fit face

on may fit side on. If the molecule enters face on, ddiagonal= 2.0 + 2×0.15

= 2.3 nm is the optimum diameter (see Figure 5.3 (b)). The same optimum

diameter would be valid if the molecule was to be rotated through 90 degrees

about an axis normal to the nanotube axis - we call this encapsulation along

the diagonal of the molecule. It is also possible for the molecule to enter

narrower nanotubes if it enters side on, in this case the optimum diameter

dside= 1.4 + 2×0.15 = 1.7 nm (see of Figure 5.3 (c)).

1.4 nm

1.4 nm

2.0 nm

NiTPP molecule

(a)

0.15 nm

2.0 nm 2.3 nm

(b)

0.15 nm

1.7 nm1.4 nm

(c)

Figure 5.3: Schematic diagrams of (a) the van der Waals surface of the NiTPP molecule(based upon [37]) and the optimum nanotube diameters for (b) face-on and (c) side-onencapsulation.

Figure 5.4 shows encapsulation of NiTPP molecules inside of nanotubes

around the optimum diameter for encapsulation along the diagonal - mea-

surements of the nanotube diameters were obtained using Gatan Digital Mi-

crograph analysis software.

81

High yield and continuous filling is observed for nanotubes with d >2.3

nm - this is consistent with a greater freedom of entry for the molecules in

these wider nanotubes. Unlike the CoPc molecules from [36] (Figure 5.2 (a)),

the NiTPP molecules appear to organize with a certain degree of ordering

forming rows of molecules showing diamond-like units. This suggests encap-

sulation of the NiTPP along its diagonal, as shown in Figure 5.4 (d). In this

case, molecules may assemble via the π-stacking of their six member ring

appendages as proposed schematically in Figure 5.4 (e).

2.0 nm

2.0 nm

2.0 nm

Figure 5.4: HRTEM of NiTPP - filled nanotubes with diameters greater than 2.3 nm, i.e.above the threshold for diagonal encapsulation. (a) Long continuous filling within a 2.3nm diameter SWNT; (b) part of the same filling as in (a) with the row of molecules shiftedtowards the center of the nanotube. (c) NiTPP inside of a 2.7 nm diameter nanotube. (d)Schematics of diagonal encapsulation of the NiTPP. (e) Proposed π-stacking of NiTPPmolecules.

Focal series (Figure 5.5) have been simulated in order to propose an as-

signment to the conformation of the NiTPP molecules inside of the nanotubes

from Figure 5.4.

82

Defocusing distance was varied in steps of 5 nm above and below the

Scherzer focus. The images at the Scherzer focus (see section 5.2) are the

ones framed in red.

Figure 5.5 (c) shows the molecule in side-on entry inside of the nanotube,

with the six-member ring appendages roughly perpendicular to the central

body of the molecule, as in free-form, while 5.5 (b) shows the appendages

being rotated to roughly align to the plane of the body of the molecule. This

rotation has been performed as none of the experimental images show the

dark contrast that accompanies the appendages positioned as in Figure 5.5

(c).

Figure 5.5: HRTEM simulated images of NiTPP entering SWNTs along diagonal (a) andside-on (b, c), respectively. Defocus varied in steps of 5 nm above and below the Scherzerfocus, located at -57 nm. Larger nanotubes are needed to accommodate the molecule whenentering along the diagonal. Parameters used for the simulation are those of the 4000HRmicroscope. In red, Ni atoms; in blue, N; in black, C; and in orange, H.

We note that the appendages of the TPP molecules are expected to be more

flexible than those of the Pc molecules (see Figure 5.2 (a)) due to just a single

C-C bond with the body of the molecule.

83

It is in this conformation of the appendages that non-metallic (un-chelated)

TPP self assembles with other TPP molecules to form J-aggregates in un-

constrained environments [75] (see Figure 5.6).

In unchelated TPP-based J-aggregates the assembly occurs through π-

stacking between the appendage of one molecule and the central body of the

neighbouring one.

Figure 5.6: Unchelated TPP self-assembled into J-aggregates (based upon [75]). The topdiagram shows sliding of the molecules along two directions in order to produce π-stacking.

However, as noted in chapter 3 section 3.3, the presence of the central

metal, as in our case, disrupts this π-bonding; the remaining possibility is π

(AB-) stacking through appendages of adjacent molecules.

Figure 5.7 demonstrates that NiTPP encapsulation has been achieved

using ScCO2 as a transporting medium in a large range of diameters, sub- and

above doptimum. Encapsulation in a sub-optimum diameter of 1.3 nm (Figure

5.7 (a)) has been observed only occasionally, therefore it has a low yield. In

addition, even when encapsulated inside of a nanotube the filling is sparse

and the molecules, seen in side view, seem highly distorted. Figure 5.7 (c)

shows that encapsulation of molecules in much larger diameters than doptimum

is not conducive of ordering, potentially due to having relaxed the spatial

constraint imposed by the nanotube template. Based upon the HRTEM

evidence gathered, it appears that ordered assembly in a single row (Figure

5.7 (b)) is highly favoured when the molecules are constrained in nanotubes

with diameter close to doptimum.

84

Figure 5.7: Experimental HRTEM images of NiTPP encapsulated inside of three typesof nanotube templates: with d < doptimum (a), (b) d ≈ doptimum, and (c) d > doptimum.

Ordering has been observed in TEM also when small rectangular perylene-

3,4,9,10-tetracarboxylic dianhydride (PTCDA) organic molecules were en-

capsulated [76]. In this case doubly stacked molecular structures were ob-

tained. The observation of the ordering in Figure 5.7 (b) might be favoured

also by the lower electron doses used in this study. In contrast, in [36] or-

dering of CoPc was not observed directly by HRTEM. It has been suggested

by [36] that the ordering may have been destroyed by interaction between

the electron beam of the TEM and the encapsulated molecules. This effect

was minimized in our work by careful choice of low electron beam doses (see

section 5.2).

85

Chapter 6

Resonant Raman spectroscopy

of filled carbon nanotubes

In this chapter Raman spectroscopy is employed to investigate the electronic

properties of SWNTs both endo-and exohedrally functionalized with planar

organo-metallic molecules. Changes in both the radial breathing mode (RBM)

and G bands are observed upon functionalisation. The changes in the posi-

tions of the nanotube G bands, known to be sensitive to charge transfer-

induced strain, are discussed in terms of charge transfer from the molecules

to the nanotube. Changes in G band peak position are also discussed in terms

of structural strain induced in the nanotube by the encapsulation process.

6.1 Introduction to the Raman effect

6.1.1 Raman-active molecules - a classical treatment

When a molecule is subjected to irradiation with an electromagnetic wave

with a frequency ν, the oscillatory electric field ~E of the electromagnetic wave

slightly changes the distribution of the electrons within the molecule. This

causes a dipole moment ~P to be induced in the molecule. ~P is proportional

to ~E and can be expressed as:

86

~P = α~E, (6.1)

where the constant of proportionality α is the polarizability of the molecule.

The polarizability can be thought of as how easily the electron cloud of the

molecule can be distorted. As both the ~E and the molecule are three dimen-

sional, α is a tensor quantity. For simplicity, we shall limit our discussion

here to one dimension. Substituting the wave equation of an oscillating elec-

tric field E = E0Cos(2πνt) into equation (6.1) the following expression is

obtained:

P = αE0Cos(2πνt), (6.2)

where E0 is the maximum amplitude of the oscillatory electric field and t

is time. In a Raman-active molecule, the polarizability of the molecule is

linked to the vibrational state of the molecule, therefore it is necessary to

add an additional term to α. The polarizability of the molecule is split into

two components, one which is independent of molecular vibration α0, and

a second which is a sum of terms having the periodic time dependence of

the normal frequencies of the system under consideration and which changes

with the molecular vibration αn. The polarizability is given by

α = α0 +∑

αncos2πνnt (6.3)

The normal frequencies νn may be the rotation or vibrational frequencies

of the system under study. Now substituting equation (6.3) into equation

(6.2) the following equation for the dipole moment of the molecule can be

obtained:

P = α0E0Cos(2πνt) +1

2E0

∑αn[Cos2π(ν −∆νn)t+ Cos2π(ν + ∆νn)t]

(6.4)

87

The first term of the expression describes the dipole moment of the

molecule oscillating with the same frequency as that the electric field of the

incident electromagnetic wave ν0. It is well known from the electromagnetic

theory of waves that an oscillating dipole emits electromagnetic radiation the

frequency of which is that of the dipole. There is no discrepancy between

the frequencies of the incident and scattered radiation, therefore this term of

the expression simply describes Rayleigh scattering.

The second term contains a component that refers to vibrations at two

different frequencies (ν −∆νn) and (ν + ∆νn), these account for the Stokes

and anti-Stokes Raman bands respectively. This equation reveals the main

pre-requisite for Raman scattering to occur - for the Raman effect to occur,

the factor∑αn must be non-zero - it means that the vibrational modes of a

molecule are Raman-active only if vibrational displacement of the molecule

results in a change in the polarizability. The practical consequence of this

selection rule is that a molecule could have a large number of possible vibra-

tional modes but only a subset of these will be Raman active [77].

The problem of describing the allowed normal modes lends itself to group

theory, with its (molecular) point groups and associated symmetry elements.

In particular, the following rule applies; if the symmetry group of a normal

mode is the same as the symmetry group of a quadratic form (x2, xy etc.)

then the mode is Raman-active [20]. For example, in the case of carbon

nanotubes, for 1st-order Raman processes, the Raman active modes have

group symmetries A, E1 and E2 [28].

6.1.2 Photonic scattering processes

While classical electromagnetic theory provides a thorough view of the Ra-

man effect, there are however, some effects that classical theory alone cannot

explain. For example, classically speaking the molecule can vibrate at any

frequency, while in reality that is not the case. A quantum mechanical treat-

ment provides a clearer view of the situation.

88

When a Raman-active material is irradiated with light of frequency ν0

a vast number of photons with quantised energy ε0 = hν0 pass through the

material unhindered. However, a fraction of the photons are reflected away

from the main direction of propagation, i.e. are scattered. The majority of

the photons undergo elastic (or Rayleigh) scattering in which a photon with

an energy ε0 incident upon the material is absorbed by the material, exciting

an electron from the ground state to a virtual state εr. There is then a rapid

decay from this state down to the ground state - a photon with a quantised

energy of εs is emitted as a result (see Figure 6.1).

ELECTRONIC GROUND STATE

0

V1

V2

V3

VIRTUAL EXCITED STATE ,Ɛr

VIBRATIONAL EXCITED STATES

RAYLEIGH STOKES ANTI-STOKES

INCIDENT PHOTON SCATTERED PHOTON

SCATTERED PHOTONINCIDENT PHOTON

INCIDENT PHOTON SCATTERED PHOTON

V

Figure 6.1: The three possible ways light can scatter from a Raman-active material(based upon [78]).

The process is elastic, therefore there is no difference ∆ε between the

energy quantra of the incident and scattering photons, such that the following

expression is true:

∆ε = ε0 − εs = 0 (6.5)

A small fraction of the scattered light undergoes Raman scattering, which

is an inelastic process. In this effect, incident photonic energy is either re-

duced or augmented by the interaction with the material. This energy change

is accounted for by a quantised electronic change in the vibrational energy

state of the molecule.

89

There are two components to Raman scattered light, one which results

from vibrational quatra ∆νn being gained by the material from an incident

photon and a second where the opposite is true - they are known as Stokes

and anti-Stokes scattering repectively.

If the material absorbs an incident photon of energy ε0 and is excited

from the ground state to a virtual state εr, the material rapidly decays from

the virtual state down to a vibrationally excited state νn, emitting a photon

with a lower energy εs in a random direction in the process; this is Stokes

scattering. The energy difference between the incident and scattered photon

is non-zero and given by the following relation:

∆ε = ε0 − εs = +∆νn, (6.6)

where ∆νn is a quantum of vibrational energy in the material.

In Stokes scattering a quantum of vibrational energy ∆νn has been im-

parted to the material, and as a result of the conservation of energy, the

scattered photon has lower energy than that of the incident. The material is

left in a vibrationally excited state.

In anti-Stokes scattering the opposite situation occurs - initially the ma-

terial exists in a vibrationally-excited state. Incident radiation excites the

material from a vibrationally excited state νn to a virtual excited state εr.

The material rapidly decays from the virtual excited state down to the elec-

tronic ground state of the material via the emission of a scattered photon with

energy εs. The anti-Stokes scattered photon is of a higher energy than that

of the incident, such that the energy difference between the two is described

by the following expression:

∆ε = ε0 − εs = −∆νn. (6.7)

In anti-Stokes scattering, the material has lost a quantum of vibrational

energy ∆νn and is left in a lower vibrational state.

90

There are different orders of Raman scattering - in 1st order Raman scat-

tering, the incident photon gives energy to a single vibration, where as in 2nd

order Raman scattering an incident photon gives energy to two vibrations.

6.1.3 A typical Raman spectrum

Materials are usually in the ground vibrational state, as a result Stokes scat-

tering occurs far more often that anti-Stokes scattering. For this reason

Stokes scattering is usually what is measured in Raman spectroscopy experi-

ments [79]. The value that is measured in a typical Raman spectrum is called

the Raman shift, this is simply the energy difference between the incident

and scattered light but in terms of wavenumber k, such that the Raman shift

is given by the following:

∆k = k0 − ks (6.8)

The energy difference ∆ε is related to the Raman shift by the ∆k by the

following equation:

∆ε =hc

2π∆k, (6.9)

where h is Planck’s constant and c is the speed of light in vacuum.

In a typical Raman spectrum, the intensity of light scattered from the

sample is plotted as a function of Raman shift. By convention, Raman shift

is measured in units of cm-1.

The Raman shift from Stokes scattered light is measured from zero up to

some maximum value imposed by limitations of the detector used to acquire

the spectrum. The Raman shift increases with wavenumber, a large Raman

shift indicates that a large amount of the energy of the incident photon has

been converted into vibrational energy in the material.

91

Every molecule or nanoscale structure has a unique set of allowed Raman-

active vibrational modes - these result in a set of bands in the Raman spec-

trum which are individual to that material. The intensity of these bands

provides information about the photon-material interaction occurring in the

sample. For example, a high intensity band is an example of a situation

where there is a strong energy exchange between the incident photons and

material.

6.1.4 Resonant Raman scattering

A powerful technique which is very useful for poorly scattering samples is

Resonant Raman spectroscopy. The resonant Raman effect occurs when the

energy of the incident light is approximately equal to an electronic transition

energy of the sample. Here the virtual excited state εr of the regular Raman

effect has been replaced by an actual excited state εn, this results in a mas-

sive increase in the intensity of the Raman scattered light. As well as the

obvious benefit of greatly increased Raman-band intensities, through careful

selection of the wavelength of the incident light the resonance effect can be

used to selectively study the substructure of large molecules [20]. With the

easy access to lasers which can provide monochromatic beams of coherent

light at a range of wavelengths, Resonant Raman spectroscopy has become

a widely used and very useful analytical technique.

6.2 Resonant Raman spectroscopy of SWNTs

Carbon nanotubes are one dimensional quantum systems, as such their elec-

tronic density of states is distributed into quantised functions called van Hove

singularities (see section 2.3.4). If an electronic transition occurs in a car-

bon nanotube, it must be from a van Hove singularity in the valence band

to a van Hove singularity in the conduction band, or vice versa, while also

obeying the selection rules.

92

The selection rules for optical transitions in carbon nanotubes depend

upon a number of aspects:

(i) The initial electronic state before the transition takes place.

(ii) The linear polarisation direction of the incident photon which will cause

the transition.

(iii) The symmetry of the phonon produced.

In general, the selection rule for optical transitions between the van Hove

singularities of the SWNT valence and conduction bands, EµV and Eµ

C re-

spectively, is given by EµV → Eµ′

C, where µ and µ′ are the initial and final

states respectively. Here µ′ = µ for incident light polarized along the axis of

the nanotube (z axis) and µ′ = µ±1 for light polarized normal to the surface

of the nanotube (x axis) [28].

When Resonant Raman spectroscopy is used to probe carbon nanotubes,

a resonance condition can be achieved when the excitation energy of the

probe laser matches that of the energy gap separating opposite van Hove

singularities. When resonance is achieved the signal from the carbon nan-

otubes is greatly enhanced. If the laser excitation energy does not match

an allowed transition energy, then no enhancement in the nanotube Raman

signal will be obtained [28].

6.2.1 Resonant Raman spectra of SWNT

The modes expected from a typical Resonant Raman spectrum are shown

below in Figure 6.2, which is an example spectrum of bundles of Nanocyl

NC1100 nanotubes. The spectrum was acquired with a laser of excitation

energy ELaser = 1.59 eV. The most prominent modes have been labelled.

93

0

10000

20000

30000

40000

50000

60000

0 500 1000 1500 2000 2500 3000

Raman shift / cm-1

Inte

nsity /

co

unts

per

seco

nd

+ SWNT bundle sample

RBM

Si D

G

M iTOLA

G'

E Laser

= 1.59 eV

G+

G-

Figure 6.2: Resonant Raman spectrum from a Nanocyl NC1100 bundle sample. Thespectrum shows the radial breathing modes (RBM), D-band, G-Band and G’ band features- the weaker M-band and iTOLA second-order modes are also observed. Signals from theoxidised silicon substrate upon which the samples sit are also present.

The most common features observed in Raman spectra of SWNTs are

listed in Table 6.1. Both first and second order Raman processes are present

in the spectrum shown in Figure 6.2 - schematic diagrams showing the pro-

cesses involved are shown in Figure 6.3.

Mode Frequency ω0 / cm-1

RBM 0 to 350D 1350G+ 1590G- 1570M 1760iTOLA 1860G’ 2700

Table 6.1: Commonly observed features in the Raman spectra of SWNTs [28].

94

The modes of both the RBM and G bands originate from a 1st order Ra-

man scattering process. In 1st order Raman scattering, one lattice phonon is

created by inelastic scattering between the crystal lattice and a laser photon.

A unique point of 1st-order scattering is that after photo-absorption, the po-

sition in reciprocal space of the excited electron (e-) k, should be the same

as the hole (h+) left behind, such that the wavevector of the phonon q = 0 -

this process is shown schematically in Figure 6.3 (a) [80].

The D, M, iTOLA and G′

modes shown in Figure 6.2 and Table 6.1 all

originate from 2nd order Raman scattering processes. In second order Ra-

man scattering, one phonon is produced, however, more than one scattering

process is in operation - electrons are scattered also.

The phonons produced in the 2nd-order processes have non-zero lattice

wavevectors, such that q 6= 0.

While q 6= 0, for Raman scattering to occur it is still necessary for the

phonon to return to point k, in order for the electron and hole to recombine

and emit a Raman scattered photon.

There are two types of second order Raman scattering, the so called 1-

phonon scattering process, where the phonon produced also elastically scat-

ters an excited electron (Figure 6.3 (b) and (c)) - such a process is responsible

for the nanotube D mode. It is possible to have both inter (Figure 6.3 (b))

or intra (Figure 6.3 (c)) valley scattering. Scattering is intra-valley if the

scattering event is restricted to the vicinity of the K or K’ lattice points, and

inter-valley if phonon scattering takes place between K and K’ or K’ and K

points [28].

95

In 2nd order scattering, two phonons are produced (Figure 6.3 (d)) [28,80].

1st Order q = 0

e-

h+

(a)

K

k

e-

h+

K K'

q

-q

2nd Order q = 0

k + q

e-

h+

K K'

q

-q

1-inelastic phonon 2nd Order q = 0

inter-valley scattering

k

k + q

(b)

(d)

e-

h+

K

k

k + q'

1-inelastic phonon 2nd Order q = 0

intra-valley scattering

(c)

q'-q'

Figure 6.3: Schematic diagrams of 1st and 2nd order resonant Raman scattering pro-cesses of carbon nanotubes. Here, k is the point in reciprocal space at which the electronexisted before being excited, the e- and h+ label the excited electron and hole created asa result of the scattering process, K and K

′are points of high symmetry in the graphene

Brillouin zone; q is the wavevector of the phonons produced during the Raman scatteringprocesses. The solid and dotted green lines represent inelastic and elastic scattering pro-cesses respectively. The black dotted line is a guide to the eye for the mid-point betweentwo inelastic scattering processes. (a) Shows first order Raman scattering, while (b) and(c) show 1-inelastic phonon second order inter and intra valley scattering respectively; (d)shows second order scattering with two inelastic phonons (based upon [80]).

Each vibrational band will now be discussed with a focus upon the use-

ful information which they can provide upon the mechanical and electrical

properties of SWNT’s. The RBM and G band features show the greatest

sensitivity to molecular doping [28], therefore a great deal of attention will

be given to them in the remainder of the chapter.

96

6.2.2 Radial breathing modes (RBM) of SWNTs - 0

to 350 cm-1

The RBM of a SWNT results from the coherent oscillation of the nanotube

side-walls, such that they appear to be breathing - a schematic representation

is shown in Figure 6.4. They are low frequency modes and occur at Raman

shifts from approximately 0 to 350 cm-1.

Radial breathing modes are extremely diameter-dependent, and the Ra-

man shift ωRBM at which they occur is inversely proportional to nanotube

diameter d such that the following expression is true [28]:

ωRBM =A

d+B, (6.10)

where A and B are constants to be determined experimentally. It has been

found that, for SWNTs with d = 1.5 ± 0.2 nm, values of A = 234 and

B = 10 cm-1 accurately predict the ωRBM at which a SWNT of d will be

located in the spectrum [28]. Here B is an up-shift to account for tube-tube

interactions due to bundling. For the usual diameter range of 1.0 > d >

2.0 nm these two parameters are accurate. However, for d < 1.0 nm the

constants become increasingly less valid, this because curvature effects begin

to have a significant effect on the SWNT properties. In addition, for SWNT

with d > 2.0 nm, accurate assignment of the ωRBM becomes increasingly

difficult as d gets larger; this is because the intensity of the RBM feature is

weak [28].

The diameter distribution of Nanocyl nanotubes used in this study is

peaked at 2.0 nm [60]; this means that the RBMs of the SWNTs at and

above the peak diameter will be difficult to observe and that the majority of

the observable RBMs will come from SWNTs on the lower diameter side of

the distribution.

If the allowed electronic transitions of the SWNTs being probed are

known, the above equation can be used to determine the diameters of SWNTs

present in a mixed sample of SWNTs and whether or not they are in reso-

nance with the probe laser.

97

Radial breathing modes of high intensity indicate that a large number of

SWNTs of a particular diameter are in resonance with the laser.

Figure 6.4: Schematic diagram of the motion of the atoms in a carbon nanotube under-going radial breathing [81].

RBM data analysis

A spectrum showing the RBM modes of a typical SWNT bundle is shown

in Figure 6.5. It can be seen from this graph that the RBMs have a low

value of the Raman shift and extend from approximately 100 to 350 cm-1 -

the low wavenumber cut-off (100 cm-1) is due to the notch filter used in the

experiments. The line-shape of an individual SWNT RBM is symmetrical

and possesses a Lorentzian character [28]. The RBM spectrum of a sample

containing many individual SWNTs can be fitted well by the superposition

of a number of Lorentzian line shapes. The typical RBM spectrum shown in

Figure 6.5 is fitted very well by a superposition of seven Lorentzian lineshapes

centred at 104 cm-1, 131 cm-1, 163 cm-1, 210 cm-1, 234 cm-1, 268 cm-1 and

305 cm-1. However, the broad peak centred at 104 cm-1 is likely to be due to

a background from Raleigh scattering. The small peak centred at 131 cm-1

may be real but its proximity to the Raleigh background makes the fitting

dubious. Using equation 6.10, the diameters of these SWNTs are calculated

to be 2.5 nm, 1.9 nm, 1.5 nm, 1.2 nm, 1.0 nm, 0.9 nm and 0.8 nm respectively.

This data fitting allows for accurate indentification of position, width and

relative intensity of the RBM’s. Both the RBM’s originating from metallic

and semiconducting SWNT’s can be fitted by Lorentzian line shapes.

98

It has been found that no significant shifts in the peak positions of RBMs

occur due to bundling of the SWNTs [82].

100 150 200 250 300 350500

1000

1500

2000

2500

3000

Data Lorentzian 1 Lorentzian 2 Lorentzian 3 Lorentzian 4 Lorentzian 5 Lorentzian 6 Lorentzian 7 Combined fit

Inte

nsity

/ co

unts

per

sec

ond

Raman shift / cm-1

Figure 6.5: The RBMs of a resonant Raman spectrum acquired from a Nanocyl NC1100bundle sample with a laser with ELaser=1.59 eV. The spectrum shows seven distinct radialbreathing modes, each fitted with a Lorentzian line-shape.

6.2.3 The G band of SWNTs - ≈ 1580 cm-1

The G mode of a SWNT is a result of the vibration of the carbon atoms that

make up the frame of the nanotube. The G mode in graphite results from

optical phonons between two dissimilar carbon atoms (i.e. one atom from

each sub-lattice see section 2.2.1) in the graphene unit cell. In graphitic

materials the G mode exhibits a single Lorentzian shaped peak at 1582 cm-1

which is related to in-plane vibrations of the carbon atoms [28].

In carbon nanotubes the picture is more complicated - instead of a single

vibrational mode there is a band formed from a number of sub-modes, the

two most intense modes are labelled G+ and G- [28]. The G+ mode results

from the vibrations of carbon atoms along the length of the nanotube and is

generally peaked at ≈ 1590 cm-1.

99

The slightly lower frequency G- mode is generally found at ≈ 1570 cm-1

and results from vibrations of carbon atoms along the circumferential direc-

tion of the SWNT - a schematic representation of the direction of vibration

of the carbon atoms in the G− and G+ nanotube modes is shown in Figure

6.6.

The lower intensity peaks of the G band, such as those located at ap-

proximately 1526 and 1606 cm-1, result from phonons belonging to symmetry

groups, such as the E2 group, that do not couple as well to the incident laser

photons.

G-

G+

Figure 6.6: Directions of vibration of carbon atoms in the G− and G+ Raman modes(based upon [83]).

100

G band analysis

In a SWNT bundle one can expect to find nanotubes of various diameters

and electronic types. Unlike in the RBM mode case, the modes of the G

band have to be fitted differently depending upon whether the metallic or

semiconducting SWNTs are present - the metallic and semiconducting G

bands are quite dissimilar in both appearance and position. Two examples

of resonant Raman spectra of individual semiconducting and metallic SWNTs

are shown in Figure 6.7.

The two most intense modes of the G band are the G- and G+ modes.

The G- and G+ modes of a semiconducting SWNT can be fitted well by two

Lorentzian lineshapes with the G+ and G- modes centred at approximately

1570 cm-1 and 1590 cm-1 respectively [28]. It can clearly be seen from the

figure that the intensity of the G+ mode is far greater than that of the G-

mode - this is a typical feature of semiconducting G bands. In order to obtain

a good fit to the entire G band of semiconducting nanotubes it is necessary

to include a number of weaker modes such as those centred at 1554 cm-1 and

1601 cm-1 [28].

Semiconducting

SWNT

1568

1592

Metallic

SWNT 1554

1588

1450 1550 1650

Frequency (cm-1)

Inte

nsity

Figure 6.7: G bands for individual semiconducting and metallic SWNTs (based upon[28]).

101

In contrast to the G band of a semiconducting SWNT, the G band of

metallic SWNTs require two dissimilar lineshapes to be used to fit the band

effectively. The metallic G+ mode is symmetrical and can be fitted by a

Lorentzian line shape centred at approximately 1588 cm-1, a clear down-shift

relative to the equivalent semiconducting peak. The G- mode is asymmetric,

tailing towards low Raman shift and is centred at approximately 1550 cm-1

- in contrast to the semiconducting G- mode, the metallic counterpart often

possesses an intensity equal to or greater than that of the G+ mode. The G−

mode can be fitted by an asymmetric Breit-Wigner-Fano (BWF) lineshape -

the BWF line shape shape results from coupling of the discrete phonons to

an electronic continuum [84]. This holds for bundled nanotubes as is the case

for the nanotubes used in this study. The presence or lack of a BWF profile

would help to discriminate from the semiconducting bands in this region.

The clear differences in line-shape and peak positions between the two

electronic types of SWNTs mean that the type of an individual tube can be

easily identified by its G band. When the two different electronic types are

mixed, as with a bundled sample, the resulting G band will be formed from

a mixture of lineshapes of the G bands of the two types. The extent to which

one type will dominate the spectrum over the other will depend upon which

is closest to resonance with the laser excitation energy. The G band of a

mixed bundle sample is shown in Figure 6.8.

A number of observations can be made from this spectrum, firstly that

the modes of both semiconducting and metallic SWNTs are present in the

G band of the bundle and the combined fit matches the experimental data

very closely. The G- (BWF fitted line) and G+ modes assigned to metallic

nanotubes are positioned at 1540 cm-1 and 1570 cm-1 respectively; the modes

assigned to semiconducting nanotubes are positioned at 1564 cm-1 and 1594

cm-1 respectively.

102

1400 1450 1500 1550 1600 16500

5000

10000

15000

20000

Data BWF Lorentzian Lorentzian Lorentzian Combined fit

Inte

nsity

/ co

unts

per

sec

ond

Raman shift / cm-1

Figure 6.8: The G band of a resonant Raman spectrum acquired from a Nanocyl NC1100bundle sample with a laser with ELaser = 2.34 eV. The spectrum shows four distinct modes,two G+ and two G- modes. The two modes belonging to metallic nanotubes (shown inmagenta) are of low intensity when compared with the dominant semiconducting modes(shown in green).

The G- (BWF fitted line) and G+ modes assigned to metallic nanotubes

are positioned at 1540 cm-1 and 1570 cm-1 respectively; the modes assigned

to semiconducting nanotubes are positioned at 1564 cm-1 and 1594 cm-1 re-

spectively. The positions of the G- and G+ modes assigned to semiconducting

nanotubes match very well with the example shown in Figure 6.7. However,

the modes assigned to metallic nanotubes do not match as well. This is most

probably due the difficulty of fitting the metallic modes to a spectrum where

the semiconducting modes are much more intense. The dominance of the

semiconducting modes implies that there are a greater number of semicon-

ducting SWNTs in resonance with the laser.

6.2.4 The D band of SWNTs - ≈ 1350 cm-1

The SWNT D band (see Figure 6.2) results from the vibration of sp3 hy-

bridized carbon atoms and as such is an indication of the amount of disorder

in the nanotubes structure. The intensity of the D mode is usually compared

to that of the G+ mode, a large value of D/G would be a indication of a

significant amount of disorder in the nanotube.

103

6.2.5 The effects of doping on the vibrational modes

of SWNTs

The SWNT Raman bands highlighted in this chapter are all sensitive to dop-

ing, with the different types of doping causing changes in either the position

or intensity of the sub-modes. There are two ways in which SWNTs are rou-

tinely doped, n-type doping where electrons are transferred from the dopants

to the SWNT and p-type doping where the opposite occurs. The other mech-

anism which can affect the Raman modes of SWNTs is the method by which

the dopant molecules are attached to the structure of the SWNT. Dopants

can be attached by either covalent bonding or in the case of aromatic dopant

molecules by π-π stacking (see section 3.1). The types of doping and the

mechanisms by which the dopants are attached to SWNTs will now be dis-

cussed in detail with a focus upon how the SWNT Raman-active modes are

affected.

Attachment mechanisms

It has been observed in a number of studies that covalent bonding between

dopant atoms or molecules and the side-walls of SWNTs resulted in a signif-

icant increase in the intensity of the SWNT D band [85–90]. This increase in

intensity has been explained in terms of a disruption to the hexagonal struc-

ture of the SWNT side-walls caused by the change from sp2 to sp3 hybridized

bonding [87,89].

In contrast to covalent bonding, π-π stacking between aromatic dopant

molecules and SWNT side-walls was found not to affect the intensity of the

SWNT D band [89].

By comparing the ratio between the SWNT D and G bands before and

after dopant molecules are bonded to SWNTs it is possible to determine

which type of bonding has occurred.

104

Doping effects

Changes in both the positions and intensities of SWNT G- and G+ have

been observed when p-type dopants were added to the sidewalls of SWNTs

[91–93]. Up-shifts in the positions of both the G- [91] and G+ modes ranging

from 3 − 9 cm-1 were observed in both semi-conducting and metallic SWNTs

upon doping [91–93]. These up-shifts in the positions of the G modes were

attributed to electron transfer from the SWNTs to the dopants [92,93]. They

have been explained in terms of a hardening or stiffening of the sp2 hybridized

bonds between the carbon atoms in the structure of the SWNTs [93].

In addition to the peak up-shifts observed in both types of SWNT, a

significant decrease in the intensity of the G- mode of metallic SWNTs was

also detected [91, 93]. In one case, where SWNTs were filled with nickel

halogenides, unfilled SWNTs showing a metallic signature were observed to

change to semiconducting upon doping with the filling [91], a clear indication

of significant charge transfer. There is also some evidence of p-type doping

resulting in decreases in the intensities of RBMs originating from metallic

SWNTs [91].

In similarity to p-type doping, changes in the position and intensities of

SWNT G modes were observed when n-type dopants were attached to the

nanotube side-walls. However, in contrast to the p-type doped SWNTs, n-

type doped SWNTs show a down shift in the positions of the G- [94–98] and

G+ [95–98] modes. These peak shifts range from 1 − 10 cm-1 and 1 − 5 cm-1

of the G- and G+ modes respectively [94,96–98].

These charge-transfer related down-shifts in peak position have been ex-

plained in terms of a weakening of the sp2 hybrized bonds between the carbon

atoms in the nanotube structure [95,96,98].

In contrast to p-type doped metallic SWNTs, the intensity of the G-

modes of n-type doped SWNTs were found to be significantly enhanced upon

doping [91, 95, 96, 98] - this again was attributed to charge transfer. The

increase in intensity has been explained in terms of a greater number of

valence electrons being available for interaction with SWNT phonons [97].

105

In conclusion, through careful analysis of the positions and intensities

of the SWNT G and RBM bands before and after doping it is possible to

determine whether charge transfer has occurred and which type of doping is

in effect.

6.2.6 Resonance conditions

In a sample of SWNTs with a range of diameters present, only those which

have allowed energy transitions matching the energy of the probe laser will

resonate.

The resonant Raman spectra shown in this chapter were acquired using a

Renishaw inVia Raman Microscope with a laser emitting light with a single

photon energy of ELaser = 1.59 eV with a maximum power of 350 mW, a

laser emitting light with a single photon energy of ELaser = 2.34 eV with a

maximum power of 350 mW and a laser emitting light with a single photon

energy of ELaser = 3.83 eV with a maximum power of 300mW, attached.

The Renishaw spectrometer was operated with a 20x microscope objective

lens when the 1.59 and 2.34 eV lasers were used; the area of the laser spot at

the point of focus on the sample for these two lasers was 50 µm2, resulting

in an energy density of 7.0 x 109 Wm-2 on the sample.

When the 3.83 eV laser was used, the Renishaw spectrometer was oper-

ated with a 40x ultra violet compatible microscope objective lens; the area

of the laser spot at the point of focus on the sample was 45 µm2, resulting

in an energy density of 6.7 x 109 Wm-2 on the sample.

The calculated energy densities assume no power loss from the laser beam

as it passes through the various optical components of the spectrometer;

however, in practice, such losses will occur. The losses will be greater for the

ultra violet light of the 3.83 eV laser due to the higher absorption of U.V. by

some of the components.

106

Laser

Adjustable stage

Sample

Mirror

Notch filter

Diffraction grating

CCD

Spectrometer

Laser beam

Light scattered from sample

Figure 6.9: A schematic diagram of the Raman spectrometer used in this study. Alaser beam is shown in green and the light back-scattered from the sample is shown multi-coloured.

The spectrometer used to acquire the Raman data analysed in this study

is shown schematically in Figure 6.9.

The beam from the laser enters the spectrometer where it is deflected by

mirrors to reach a beam-splitter - this deflects majority of the light from the

beam through the microscope of the system to arrive at the sample which

sits upon an adjustable stage.

Light which is back-scattered by the sample (both Rayleigh and Raman

scattering) enters the microscope, passes back through the beam splitter and

is then incident upon a notch filter - the notch filter attenuates greatly the

comparatively intense Rayleigh-scattered light and lets through the lower

frequency Raman-scattered light.

After passing through the notch filter, the spectral frequencies of the

remaining scattered light are separated out by a diffraction grating - this

light in turn falls upon a charge coupled device (CCD) where the intensities

of the individual spectral frequencies are recorded. The data from the CCD

is processes by software installed on the PC of the system.

107

Electronic considerations

The allowed transition energies in SWNTs are primarily dependent upon

diameter. Using the general expression for the energy dispersion for SWNTs

(see section 2.3.3) it is possible to calculate the allowed electronic transition

energies of individual SWNTs using their n and m numbers and plot them as

a function of nanotube diameter. Such a graph is known as a Kataura plot.

An example Kataura plot [99] is shown in Figure 6.10 (a). In this plot the

allowed transition energies of a large number of SWNTs have been plotted

as a function of diameter.

A Kataura plot covering the range of diameters and laser excitation en-

ergies used in a study is very useful in enabling one to identify the electronic

type (i.e. metallic or semiconducting) of a SWNT with a certain diameter

which is in resonance with particular laser excitation energy.

The carbon nanotubes used in this study are Nanocyl NC1100 SWNTs

[60]. They were grown using a carbon vapour deposition (CVD) technique

and have an average diameter of 2.0 nm [60] - an example diameter distribu-

tion for these nanotubes is shown as a blue curve on the modified Kataura

plot shown in Figure 6.10 (b). This curve is based upon a Gaussian function

fitted to the diameter distribution of the CVD grown SWNTs of reference

[61]. When used in conjunction with a Kataura plot like that shown in Figure

6.10 (a) it is possible to identify the electronic types of the SWNTs likely

to be in resonance with a particular laser by matching up the nanotube di-

ameters (d) with the photon energy of the laser ELaser. However, it is also

necessary to take the bundled state of the nanotubes into account. Calcula-

tions have shown that the van Hove singularities of SWNTs are shifted by

as much as 100 meV upon aggregation into bundles [100]. Practically, this

is observed as a red-shift in the absorption spectra of bundles of nanotubes

[101]. This means that the allowed transition energies of individual SWNTs

shown in the Kataura plot (Figure 6.10 (a)) are 0.1 eV higher than what

would be expected for bundled nanotubes - this has been taken into account

in the modified Kataura plot shown in Figure 6.10 (b).

108

This will affect which nanotube diameters are in resonance with the laser,

although it can be seen from the Kataura plot that the electronic type (semi-

conducting or metallic) of nanotube in resonance will be the same.

In addition to the effects of nanotube bundling, it is also necessary to

include the nanotubes which are in resonance with the outgoing photon scat-

tered from the nanotube G modes, of energy Eii = ELaser - Ephonon, with

Ephonon ≈ 0.2 eV [102]. Bands have been added to the modified Kataura plot

shown in Figure 6.10 (b) to account for this.

Figure 6.10: (a) The allowed electronic transition energies Eii vs. nanotube diameter dfor SWNTs calculated using the nearest-neighbour tight binding method, with the trans-fer integral γ0 = 2.9 eV , the carbon-carbon distance aC-C = 0.144 nm, and neglectingnanotube curvature effects (based upon [99]). The superscript of the energy, E, refersto either semiconducting (S) or metallic (M) nanotubes, while the subscript refers to theelectronic transition from the initial state (i) to a symmetric final state (i). The closedblack and the open red circles indicate semiconducting and metallic SWNTs respectively.The red and green lines indicate the excitation energies of the red and green lasers usedto probe the samples. (b) In this diagram, the plot has been shifted down by 0.1 eV totake into account the increase in the energy of the electronic transitions due to nanotubebundling. The laser lines have been replaced by bands to represent the possibility of res-onant scattering between the nanotube electronic transitions and the Raman scatteredphotons from the G modes. A curve based upon a Gaussian fit to the CVD nanotubes of[61], has been included in blue to indicate the nanotube diameters which will contributemost to resonance.

109

The photon energies of the red (ELaser=1.59 eV) and green (ELaser= 2.34

eV) lasers used in this study have been added to Figure 6.10 with the

appropriate colours.

(i) ELaser = 1.59 eV It can be seen that with a d ≈ 2.2 nm the ES33 semi-

conducting allowed transition energy will be in resonance with the red

laser. With this in mind, one would expect the resonant Raman spec-

tra acquired using this laser to have a predominantly semiconducting

character. However, it is likely that there may also be a contribu-

tion from the EM11 allowed transition originating from metallic SWNT

with diameters of ≈ 1.7 nm which might lend the spectra some metallic

character.

(ii) ELaser = 2.34 eV The photon energy of the green laser is greater than

that of the red laser. At this higher energy the SWNT allowed tran-

sition energy bands are much more tightly bunched. It can be seen

from the figure that the resonance window is centred between the EM22

metallic and ES44 semiconducting transitions and therefore it is likely

that the Raman spectra acquired using this laser will have a mixed

character. Due to the tight bunching of the allowed transitions it is

likely that significant contributions will also originate from the ES44

and ES33 semiconducting allowed transitions of SWNTs with diame-

ters of slightly less than or greater than 2.0 nm. With this in mind one

would expect resonant Raman spectra acquired with a green laser to

have a mixed character with contributions from both semiconducting

and metallic SWNTs, but the semiconducting character will probably

dominate.

(iii) ELaser = 3.83 eV A U.V. laser with a photon energy of 3.83 eV was also

used in this study. This energy is beyond the range of the Kataura plot

shown in Figure 6.10, however, SWNT transitions exist at this energy.

At this energy the SWNT electronic transitions are even more tightly

packed. As a result it is likely that in similarity to the situation with

the green laser the resonant Raman spectra acquired using the U.V.

laser will be composed of contributions from both semiconducting and

metallic SWNTs.

110

Modal considerations

When considering the expected resonant Raman spectra one would expect to

acquire for a SWNT sample containing a range of diameters it is necessary

to bear in mind the resonance of the individual bands.

For example, in the case of the SWNTs used in this study, the largest

contribution to the Raman G bands will originate from SWNTs with a di-

ameter of approximately 2.0 nm. Intuitively, one might expect the same to

be true for the RBM band, however, this is not the case. It was mentioned

in section 6.2.2 that the RBMs of SWNTs with diameters of greater than

2.0 nm do not resonate well [28], with the intensity of the mode varying in-

versely to nanotube diameter. Therefore, the strongest modes will originate

from SWNTs with d < 2.0 nm.

6.3 Experimental considerations

Nanocyl SWNTs were functionalized both endohedrally using the ScCO2

method, and exohedrally using a solution mixing method (see chapter 4,

section 4.6). The molecules used to form the SWNT/molecules samples are

those described in chapter 4, section 4.6.

These two methods were utilised to produce powders of SWNTs filled

and covered with molecular systems respectively. After a washing process

to remove any excess molecules (see chapter 4, section 4.7) the hybrids of

SWNTs/molecules samples were drip condensed onto SiO2/Si substrates pro-

ducing mat samples.

In this study evidence of charge transfer from the dopant molecules at-

tached to the interior and exterior surface of the SWNTs has been sought

from changes in the peak positions and relative intensities of the SWNT

Raman-active modes. As such it was first necessary to investigate and dis-

criminate the possible environmental sources which could give rise to such

changes in the peak positions and hence confuse charge transfer-induced

changes.

111

6.3.1 Environmental effects upon the resonant

Raman spectra of SWNTs

In this section the environmental effects which can cause perturbations to

the resonant Raman spectra of SWNTs will be discussed. There are three

main environmental effects which can cause varying degrees of change to the

resonant Raman spectra of SWNTs, they are: (i) effects of contact with the

substrate, (ii) thermal effects and (iii) the effects of the vibrational modes of

foreign systems. The latter will be discussed in a separate section.

6.3.2 (i) Effects of contact with the substrate

It has been observed at the individual SWNT level that contact with a sub-

strate, for example silicon, can cause strain induced changes in the resonant

Raman spectrum of the nanotube. This is seen as an up-shift of both the D

and G bands of the SWNT, and a modification of line-shape of the G band.

The RBM is unaffected [103].

The SWNT samples used in the present study are bundled and form a

layer of a significant thickness on top of a SiO2/Si substrate. The bundled

nature and layer thickness make such contact induced strain effects highly

unlikely and bundling will be the dominant effect.

6.3.3 (ii) Thermal effects

The environmental factor which has the most significant effect on the res-

onant Raman spectra of SWNTs results from fluctuations in temperature.

There are a number of ways in which the temperature of a SWNT sample can

induce perturbations to the acquired spectra. The most commonly reported

is a downshift with increasing temperature in the radial breathing modes, D

and G modes [102–111].

This effect has been observed in the spectra of SWNTs heated by a laser

and by conventional means [105].

112

Indeed, upon laser illumination of the SWNT samples used in this study

with increasing laser power, down-shifts in the peak positions of the radial

breathing modes and G modes were observed. The extent of the observed

down-shift was directly related to the power of the laser. Radial breathing

modes and G modes acquired using a ELaser = 2.34 eV laser set to a range

of powers are plotted in Figure 6.10.

It can be seen from the spectra that the down-shift in peak position of

both the radial breathing modes and G modes become increasingly large as

the laser power is increased from 10 % to 50 % of the maximum power. It

can also be seen that the maximum down-shift is more pronounced in the G

mode (22 cm-1) than the RBMs (3 cm-1). The down-shifts in peak position

have been attributed mainly to the weakening of the C-C bonds between

the carbon atoms of the SWNT and weakening of the van der Waals bonds

between the SWNTs forming bundles [108].

Spectra acquired from samples after they had been left to cool were found

to exhibit some differences compared to the spectra acquired before the heat-

ing series was conducted. Changes were observed in the relative shapes and

intensities of both the RBM and G bands, together with the appearance of

additional modes in the RBM band.

The other way in which heating effects can perturb SWNT spectra is by

introducing variations in the relative peak intensities of the RBMs [104,106,

107, 109]. This effect has been observed in Figure 6.11 - as the power of the

illuminating laser was increased the relative intensities of the RBMs change

and in some instances, new modes appear.

113

100 150 200 250 3000

2000

4000

6000

8000

10000

12000

14000

16000

10% power 30% power 50% power

Inte

nsity

/ co

unts

per

sec

ond

Raman shift / cm-1

(a)

1400 1500 16000

20000

40000

60000

80000

100000

120000

140000

10% power 30% power 50% power

Inte

nsity

/ co

unts

per

sec

ond

Raman shift / cm-1

Peak down-shift

1584 cm-1

1591 cm-1

1569 cm-1

(b)

Figure 6.11: Resonant Raman spectra of unfilled SWNTs acquired using a ELaser = 2.34eV laser set to 10%, 30% and 50% of the maximum power. (a) and (b) show the RBMand G bands respectively. The sloping magenta lines indicate down-shifts in peak positionand the green box shows a RBM mode which has been enhanced as a result of heating.

114

This heating effect has been attributed to thermally induced changes in

the electronic density of states of the SWNTs [104, 106, 107, 109]. It has

been explained in terms of the allowed transition energies of the SWNTs,

Eii, coming into and out of resonance with the energy of the laser, ELaser. For

example, if Eii and ELaser are in resonance, a strong RBM signal is observed.

In contrast, if the SWNT is heated causing Eii to decrease, the resonance with

the laser can be broken and in the extreme case the mode will disappear from

the spectrum. The opposite can also occur, with heating modifying Eii in

such a way that it is brought into resonance with ELaser, thus causing a mode

which was previously out of resonance to start to resonate, introducing a new

mode to the spectrum.

Another factor which can affect the amount of peak shift is the thermal

conductivity of the sample. At the individual SWNT level it has been ob-

served that intimate contact between the SWNT and a SiO2/Si substrate

provides a sufficient thermal contact such that the substrate acts as a heat-

sink for the SWNT and no peak down-shifts are observed. However, when

the same measurements were acquired on an unsupported part of the same

nanotube, thermally induced peak down-shifts, similar to those observed in

bulk samples, were observed [103]. Clearly the thermal conductivity of the

sample plays a big part in determining the effectiveness of the substrate as

a heat-sink.

It has been observed that the amount of disorder in the sample has a sig-

nificant effect upon the thermal conductivity of carbonaceous samples [105].

For example, if a SWNT sample contains a large amount of amorphous car-

bon the ability of the sample to conduct laser-imparted heat to the substrate

is severely compromised. In bundled samples, it might be advantageous to

acquire spectra from areas of the sample where the SWNT layer is thin. This

might maximise the heat transmission to the substrate and minimise the heat

stored in the body of the sample.

Another way in which heating effects can be minimised is by placing the

sample in an aqueous solution, hence providing liquid cooling [103].

115

This however, is deemed too risky for powder samples, primarily due to

the possibility of the solution lifting the SWNT mat off the substrate, thus

ruining the sample; in addition, the Raman spectra of nanotubes have been

observed to change when immersed in water [112].

In conclusion, out of the possible environmental effects which can cause

perturbations to the resonant Raman spectra of SWNTs it is the tempera-

ture effects which are by far the most significant. They have been found to

affect both the positions of the radial breathing, D and G modes and the rel-

ative intensities of the RBMs in bundled samples. It is therefore imperative

to avoid heating effects if one is to attribute changes in peak position and

intensity to non-thermal effects such as charge transfer. There are a number

of ways in which the risk of heat-induced shifts can be minimised. The most

obvious and effective is to use a laser set to a relatively low power coupled

with a low magnification objective lens which will spread the laser spot over

a larger area.

The next section will focus upon the experimental procedures which were

developed to eliminate undesirable thermal effects.

6.3.4 Heating control experiments

In order to be certain that any changes in the SWNT spectra were due to

SWNT-molecule interactions, it was first necessary to rule out the influence of

heating effects. The procedure by which this was achieved is described below.

It was mentioned in section 6.3.3 that the position of the G band is

very sensitive to heating, therefore it presents the ideal mechanism by which

heating can be detected. In order to determine the position of the G band

in the absence of heating effects, a resonant Raman spectrum was acquired

from an unmodified SWNT sample using a 2.34 eV laser set to a very low

power of 0.1% of the maximum laser power (350 mW) and a short acquisition

time of 1 second. It can be seen from the spectrum acquired (Figure 6.12)

that the G peak position is 1596 cm-1 - this agrees with the expected peak

position of a semiconducting SWNT.

116

1400 1500 16000

20

40

60

80

SWNTs 0.1% power

Inte

nsity

/ co

unts

per

sec

ond

Raman shift / cm-1

1596 cm-1G+

Figure 6.12: Resonant Raman spectra of unfilled SWNTs acquired using a 2.34 eV laserset to 0.1 % of the maximum laser power with an acquisition time of 1 second.

In order to determine the largest intensity that could be used without

causing heating a series of readings was taken in which the sample was illu-

minated by the laser for 1 second at 1% and 5% of the maximum power - the

results are shown in Figure 6.13. Comparing the positions of the G+ modes

of the spectra acquired with 1% and 5% maximum power with the spectrum

acquired using 0.1% maximum power (Figure 6.12), it can be seen that a

large peak down-shift of 22 cm-1 was observed when 5% maximum power

was used (a clear indication of laser-induced heating) and a small down-shift

of 1 cm-1 was also observed when 1% maximum power was used.

117

1400 1500 16000

100

200

300

400

500

G-

G+

SWNTs 1% power

Inte

nsity

/ co

unts

per

sec

ond

Raman shift / cm-1

1595 cm-1

(a)

1400 1500 16000

500

1000

1500

2000 G+

SWNTs 5% power

Inte

nsity

/ co

unts

per

sec

ond

Raman shift / cm-1

1574 cm-1

(b)

Figure 6.13: Resonant Raman spectra of SWNTs acquired using a 2.34 eV laser set to(a) 1 % and (b) 5 % maximum power and an acquisition time of 1 second.

With some fine tuning of the laser power, it was found that the highest

power which did not result in a peak shift was 0.5% maximum power. An

example spectrum acquired with 0.5 % maximum power for 1 second is shown

in Figure 6.14 (a). This spectrum is a lot more intense than the spectrum

acquired with 0.1 % maximum power (Figure 6.12), however, the signal to

noise ratio is still not as high as desirable.

118

To improve the signal to noise ratio of the spectra, longer acquisition

times were experimented with. A number of different acquisition times were

tried ranging from 1 second to a few minutes. It was found that a good signal

to noise ratio was obtained when an acquisition time of 30 seconds was used

(Figure 6.14 (b)). It can be seen from this spectrum that there is a significant

increase intensity of the G band in the 30 seconds spectra compared to that

acquired using a 1 second acquisition time and the signal to noise ratio is

greatly improved.

1400 1450 1500 1550 1600 16500

100

200

300

400

500

G-

G+

SWNTs 0.5% power 1 second exposure

Inte

nsity

/ co

unts

per

sec

ond

Raman shift / cm-1

1596 cm-1

(a)

1400 1450 1500 1550 1600 16500

2000

4000

6000

8000

10000

G-

G+

SWNTs 0.5% power 30 seconds exposure

Inte

nsity

/ co

unts

per

sec

ond

Raman shift / cm-1

1596 cm-1

(b)

Figure 6.14: Resonant Raman spectra of SWNTs acquired using a 2.34 eV laser using0.5 % maximum power with an acquistion time of (a) 1 second and (b) 30 seconds.

119

To check whether a longer exposure of the sample to the laser beam would

produce a noticeable temperature change in the sample, a number of expo-

sure series were acquired on different positions on the sample. An exposure

series consisted of illuminating the sample continuously for 30 seconds and

acquiring a spectrum every second. This allowed for the temporal evolution

of the sample under laser irradiation to be investigated. A representative

time series acquired using 0.5 % maximum power is shown in Figure 6.15 -

a time series acquired using 5% maximum laser power is shown for compari-

son. It can be seen from this figure that the position and intensity of the G+

mode in the component spectra of the series acquired using 0.5 % maximum

power remain constant at a value of 1596 cm-1 over time, this implies that

no significant heating effects have occured.

In contrast, the component spectra of the series acquired with 5 % max-

imum power is temporally dependent. Two main observations can be made

from this graph - one that the intensity of the SWNT G band increases with

time, as such it was not necessary to displace the spectra for clarity as in

Figure 6.15 (a). Secondly, the down-shifted G+ mode centered at 1574 cm-1

shows a small up-shift with time. These observations indicate that laser-

induced heating is causing modification the sample with time. The fact that

the changes are beneficial to the signal strength of the G band implies that

the laser induced heating is burning off some of the impurities present in

the sample. Such a conclusion would be consistent with more light reaching

the SWNTs and hence result in a greater signal strength and fewer impuri-

ties resulting in a higher thermal conductivity of the sample resulting in a

cooling-induced up-shift in the G band.

120

1400 1450 1500 1550 1600 1650

0

200

400

600

800

1000

1200

G-

G+

After 1s After 5s After 10s After 15s After 20s After 25s After 30s

Inte

nsity

/ co

unts

per

sec

ond

Raman shift / cm-1

1596 cm-1

(a)

1400 1450 1500 1550 1600 16500

1000

2000

3000

4000

5000 G+

After 1s After 5s After 10s After 15s After 20s After 25s After 30s

Inte

nsity

/ co

unts

per

sec

ond

Raman shift / cm-1

1574 cm-1

(b)

Figure 6.15: Resonant Raman spectra of SWNTs acquired using a 2.34 eV laser set to0.5 % (a) and 5% (b) maximum power. Each coloured trace on the graphs represents aone second acquisition from the sample, with 30 being acquired from the sample for eachlaser power. The individual spectra shown in Figure (a) have been artificially separatedalong the intensity axis for clarity.

121

Having determined that no heat-induced changes occur after a 30 second

exposure with the laser set to 0.5 % maximum power, the final effect to test

was whether variations in the thickness of the SWNT layer of the sample

produced noticeable changes in peak position or shape. This was probed by

acquiring spectra at a number of different locations on the sample using the

laser set to 0.5 % maximum power with an acquisition time of 30 seconds -

some representative spectra are shown in Figure 6.16.

It can be seen from Figure 6.16 that there is little or no variation in

the Raman shift of the SWNT G bands - this implies that the power and

acquisition time are such that they do not cause position-sensitive effects to

appear in SWNT spectra.

Similarly, the radial breathing modes do not show any noticeable change

in the positions of the most intense modes; however, there are a number

of minor changes to the lines-shape as one moves from position to position.

This is most likely to be a result of variations in the types and diameters

of the SWNTs which form the bundles in the sample. The changes in the

intensity of the modal peaks are likely due to either variations in the density

of the samples in different regions of the sample or differences in the focus of

the laser.

In conclusion, SWNT resonant Raman spectra which were free from

heating-induced perturbations were acquired using a 2.34 eV laser set to

0.5 % maximum power and set to acquire for 30 seconds.

122

50 100 150 200 250 3000

200

400

600

800261 cm

-1182 cm

-1152 cm

-1

Position 1 Position 2 Position 3

Inte

nsity

/ co

unts

per

sec

ond

Raman shift / cm-1

102 cm-1

(a)

1400 1450 1500 1550 1600 16500

2000

4000

6000

8000

10000

12000

14000

16000

G-

G+1596 cm-1

Position 1 Position 2 Position 3

Inte

nsity

/ co

unts

per

sec

ond

Raman shift / cm-1

(b)

Figure 6.16: Resonant Raman spectra of unmodified SWNTs acquired using a 2.34 eVlaser set to 0.5 % of the maximum laser power with an acquisition time of 30 secondsacquired from three different positions on the sample, (a) and (b) show the RBM and Gbands of the spectra respectively.

The same procedure as outlined above was used to determine the optimum

conditions for the 1.59 eV and 2.34 eV lasers.

123

SWNT spectra which were free from heating effects were acquired when

the 1.59 eV laser was set to 0.5 % and 0.1 % of the maximum power (350 mW)

depending upon whether the system was operated in laser spot or laser line

mode respectively; and when the 3.83 eV laser was set to 5 % of the maximum

power (300 mW). As discussed in section 6.2.6, the most likely reason for the

need for a higher power setting for the U.V. laser is to compensate for the

extra absorption of U.V. by some of the components of the spectrometer.

6.3.5 (iii) Vibrational modes of non-nanotube

components

Another possible cause of perturbation to the SWNT spectra is contributions

from the presence of vibrational modes originating from molecules and struc-

tures sharing the environment of the SWNTs. In the experiments conducted

in this study the two main sources are the SiO2/Si substrates upon which the

samples were deposited and the molecular species with which the SWNTs

were filled or covered.

Vibrational modes of silicon

Raman spectra were acquired from a SiO2/Si substrate which is represen-

tative of those used in this study using 2.34 eV and 1.59 eV lasers and are

are shown in Figure 6.17 below. It can be seen from parts (a) and (b) of

the figure that the Raman spectra acquired from the silicon substrate using

both laser wavelengths show sharp intense modes at 521 cm-1. This is a well

known vibrational mode and was used to calibrate the spectrometers used in

this study. Given the sharpness and position of this mode it is unlikely that

it would cause any problems in SWNT peak identification.

124

500 1000 1500

0

2000

4000

6000

8000

10000

12000

14000

16000

Si substrate - 2.34 eV laser

Ite

nsity / c

ou

nts

per

se

con

d

Raman shift / cm-1

521 cm-1

Raman shift / cm-1

500 1000 1500

0

2000

4000

6000

8000

10000

12000

14000

16000

Si substrate - 1.59 eV laser

Inte

nsity / c

ounts

per

seco

nd

Raman shift / cm-1

521 cm-1

Raman shift / cm-1

500 1000 1500

0

10000

20000

30000

40000

A typical SWNT bundle - 2.34 eV laser

Inte

nsity / c

ounts

per

seco

nd

Raman shift / cm-1Raman shift / cm-1

500 1000 1500

0

50000

100000

150000

200000

A typical SWNT bundle - 1.59 eV laser

Co

un

ts p

er

se

con

d

Raman shift / cm-1Raman shift / cm-1

(b)

(c) (d)

(a)

Inte

nsity / c

ounts

per

seco

nd

Figure 6.17: Raman spectrum of silicon acquired with a 2.34 eV energy laser (a) and a1.59 eV energy laser (b). (c) and (d) show resonant Raman spectra from SWNTs placedon the SiO2/ Si substrate acquired with a 2.34 eV and 1.59 eV energy laser respectively.

Resonant Raman spectra were acquired from a SWNT sample which is

representative of those used in the study using 2.34 eV and 1.59 eV lasers and

are shown in parts Figure 6.17 (c) and (d) respectively. It can be seen that

there is no evidence of the relatively intense 521 cm-1 Si vibrational mode

in either of the spectra. This implies that the thickness of the SWNT layer

in the samples is of sufficient thickness so that little or no light reaches Si

substrate upon which it is deposited. It is therefore reasonable to conclude

that providing there is a sufficient covering of SWNTs, perturbations to the

SWNT spectra from the substrate are unlikely.

125

The vibrational modes of molecular systems

The other source of vibrational modes which could disrupt the SWNT modes

originate from the molecular species with which the SWNTs have been filled

or covered. Raman spectra were acquired from the molecules used in this

study using a 2.34 eV and 1.59 eV lasers and are shown in Figure 6.18.

It can be seen that clear spectra with a high degree of spectral resolution

were acquired using the 2.34 eV laser. The spectral clarity is such that the

vibrational modes of the molecules can be clearly identified, meaning that

any molecule peaks which appear in any of the modified SWNT spectra can

be easily identified - the most intense modes have been labelled for clarity.

It can be seen from the spectra shown in parts (b) and (c) of Figure 6.18

that the phthalocyanine (Pc) molecules possess intense vibrational modes in

the region of 1520 - 1545 cm-1 - it is these intense modes which are most

likely to cause confusion in the interpretation of the SWNT G band. The

vibrational modes of metallic phthalocyanines at this frequency have been

attributed to stretches in the C-N-C bonds as well as expansion of the pyrrole

structure coupled with C-H in-plane bending (IPB) vibrations in the molecule

[113]. The position of this peak varies depending upon which metal ion is

present at the core of the molecule. The Raman spectrum of the NiTPP

molecule shown in part (a) of the figure show that there are no high intensity

molecular vibrational modes present which might overlap with the SWNT G

band.

126

500 1000 1500

0

2000

4000

6000

8000

10000

12000

NiTPP molecule - 2.34 eV laser

Inte

nsity / c

ounts

per

seco

nd

Raman shift / cm-1

500 1000 1500

0

10000

20000

30000

40000

50000

60000

NiTPP molecule - 1.59 eV laser

Inte

nsity / c

ounts

per

seco

nd

Raman shift / cm-1

500 1000 1500

0

50000

100000

150000

200000

ClAlPc molecule - 1.59 eV laser

Inte

nsity / c

ounts

per

seco

nd

Raman shift / cm-1

500 1000 1500

0

5000

10000

15000

20000

25000

30000

NiPc molecule - 2.34 eV laser

Inte

nsity / c

ounts

per

seco

nd

Raman shift / cm-1

1545 cm-1

500 1000 1500

0

2000

4000

6000

8000

10000

ClAlPc molecule - 2.34 eV laser

Co

un

ts p

er

se

con

d

Raman shift / cm-1

1520 cm-1

500 1000 1500

0

5000

10000

15000

20000

25000

NiPc molecule - 1.59 eV laser

Inte

nsity / c

ounts

per

seco

nd

Raman shift / cm-1

1548 cm-1

Raman shift / cm-1Raman shift / cm-1

Raman shift / cm-1 Raman shift / cm-1

Raman shift / cm-1 Raman shift / cm-1

Inte

nsity /

co

un

ts p

er

se

co

nd

(a) (b)

(c) (d)

(e) (f)

Figure 6.18: (a), (b) and (c) show Raman spectra of the NiTPP, ClAlPc and NiPcmolecules respectively, acquired using a 2.34 eV laser. (d), (e) and (f) show the equivalentspectra, but acquired using a 1.59 eV laser. The molecules from which these Ramanspectra were acquired were deposited onto SiO2 / Si substrates.

127

Figure 6.19 (a) and (b) show a comparison between spectra acquired from

a sample of un-modified SWNTs and of SWNTs with a surface covering of

NiPc molecules. This particular molecular species was chosen because it was

seen to provide the greatest overlap with the SWNT spectrum. It can be

seen from these spectra that there are clearly some vibrational modes in

the doped SWNT spectrum which are not present in the undoped spectrum,

which can be attributed to the NiPc molecule. However, it can be seen that

the SWNT spectrum is clearly still very dominant.

500 1000 1500

0

2000

4000

6000

8000

10000

12000

14000

Unmodified SWNTs - 2.34 eV laser

Inte

nsity / c

ounts

per

seco

nd

Raman shift / cm-1

500 1000 1500

0

20000

40000

60000

80000

100000

120000

140000

Unmodified SWNTs - 1.59 eV laser

Inte

nsity / c

ounts

per

seco

nd

Raman shift / cm-1

Raman shift / cm-1

Raman shift / cm-1

(a) (b)

(c)

500 1000 1500

0

5000

10000

15000

20000

25000

30000

SWNTs covered with NiPc molecules - 2.34 eV laser NiPc molecule - 2.34 eV laser

Inte

nsity / c

ounts

per

seco

nd

Raman shift / cm-1Raman shift / cm-1

500 1000

200

400

600

800

Inte

nsity /

co

un

ts p

er

se

co

nd

Raman shift / cm-1

500 1000

Raman shift / cm-1

Inte

nsity / c

ounts

per

seco

nd

200

400

600

800

500 1000 1500

0

20000

40000

60000

80000

100000

NiTPP covered SWNTs - 1.59 eV laser

Inte

nsity / c

ounts

per

seco

nd

Raman shift / cm-1

Raman shift / cm-1

(d)

500 1000

2500

5000

7500

10000

Inte

nsity /

co

un

ts p

er

se

co

nd

Raman shift / cm-1

Raman shift / cm-1

500 1000

2500

5000

7500

10000

Inte

nsity / c

ounts

per

seco

nd

Figure 6.19: Raman spectra of (a) unmodified SWNTs and (b) SWNTs covered withNiPc molecules acquired using a 2.34 eV laser. The Raman spectrum of the NiPc moleculeshas been included in (b) for comparison. Areas of the modified SWNTs Raman spectrumwhich have been modified by the presence of the NiPc spectrum are highlighted in greenboxes. The inset shows a zoomed in version of the Raman spectrum of SWNTs coveredwith NiPc molecules acquired using a 2.34 eV laser. (c) and (d) of the figure show Ramanspectra of unmodified SWNTs and SWNTs covered with NiTPP molecules respectively-the Raman spectra were acquired using a 1.59 eV laser. The inset in (d) shows a zoomedin version of the Raman spectrum of SWNTs covered with NiTPP molecules acquiredusing a 1.59 eV laser.

128

When illuminated with a 1.59 eV laser both the NiTPP and ClAlPc

molecules showed very large and intense fluorescence peaks (see Figure 6.18

(d) and (e)). These could cause a significant overlap with the SWNT spectra

if a sufficient amount of molecules were present. In Figure 6.19 (c) Raman

spectra acquired using a 1.59 eV laser from a sample of unmodified SWNTs

and Figure 6.19 (d) SWNTs modified with NiTPP molecules are compared.

It can be seen that there is little or no difference between the SWNT spectra

before and after modification with NiTPP molecules, specifically, no large

fluorescence features are observed, and the SWNT modes clearly dominate

the spectrum.

6.4 Results and discussion

6.4.1 The SWNT G band

2.34 eV and 3.83 eV lasers have been used to probe both metallic and semi-

conducting nanotubes, while the 1.59 eV laser will probe predominately semi-

conducting nanotubes. The different laser excitation energies will only res-

onate with nanotubes specific diameters.

The component modes of the SWNT G band from nanotubes probed with

each laser, show changes in both the positions and relative intensities of the

G+ and G- modes upon modification of the nanotubes with organo-metallic

molecules.

Changes in the peak position of these modes have been investigated to

determine whether charge transfer between nanotube and dopant molecules

had occured.

Spectra acquired using the 1.59 eV laser

The spectra acquired using the 1.59 eV laser were fitted well by four Lorentzian

line-shapes centred at approximately 1546 cm-1, 1566 cm-1, 1593 cm-1 and

1601 cm-1.

129

1400 1450 1500 1550 1600 1650

ELaser= 1.59 eV Unmodified SWNTs SWNTs filled with NiTPP molecules SWNTs covered with NiTPP molcules

Inte

nsity

/ A.

U.

Raman shift / cm-1

(a)

1400 1450 1500 1550 1600 1650

ELaser = 1.59 eV Unmodified SWNTs SWNTs filled with ClAlPc molecules SWNTs covered with ClAlPc molecules

Inte

nsity

/ A.

U.

Raman shift / cm-1

(b)

1400 1450 1500 1550 1600 1650

ELaser= 1.59 eV Unmodified SWNTs SWNTs filled with NiPc molecules SWNTs covered with NiPc molecules

Inte

nsity

/ A.

U.

Raman shift / cm-1

(c)

Figure 6.20: (a) to (c) of the figure show resonant Raman spectra of SWNTs filled andcovered with NiTPP, ClAlPc and NiPc molecules respectively, acquired using a 1.59 eVlaser. The spectra of the modified SWNTs for each molecular species have been normalizedto the G+ mode of the unmodified SWNT spectrum for clarity. The spectra of unmodifedSWNTs are shown in black as a reference.

130

The vibrational modes located at 1566 cm-1 and 1593 cm-1 are consistent

with the G- and G+ vibrational modes of semiconducting SWNTs. This is

in agreement with the resonance conditions described in section 6.2.6.

In order to achieve a good fit to the experimental data it was necessary to

include some of the weaker modes of the SWNT G band, specifically those

located at 1546 cm-1 and 1601 cm-1. These modes are not thought to be

sensitive to charge transfer, variation in the fitted peaks assigned to these

modes are most likely an artefact of the data fitting process. In addition,

given the low amplitude of these peaks and the lack of any well defined

signatures in the G bands of the acquired spectra, it is reasonable to put

a low weighting to the significance of the changes in the positions of these

peaks and instead to focus upon the G- and G+ modes. The resonant Raman

spectra acquired are shown in Figure 6.20.

It can be seen from the fitted peak positions shown in Table 6.2 that

there are small up-shifts of 1 cm-1 in the positions of the SWNT G- and G+

modes upon being filled with NiTPP molecules. There is also a decrease in

the relative intensity of the G- peak relative to the G+ peak. The SWNTs

covered with NiTPP molecules also show a small up-shift of 1 cm-1 in the

position of G- mode, however no shift in the G+ mode is observed and there

is negligible change in the relative intensity of the G- mode.

A comparison of the results from SWNTs filled and covered with NiTPP

molecules shows that the greatest changes to the SWNT vibrational modes

occur when the nanotubes are filled with molecules.

In contrast to the situation observed when SWNTs were functionalized

with NiTPP molecules, down-shifts in the positions of the G+ and G- peaks

are seen when SWNTs were modified with ClAlPc molecules.

These down-shifts are significant, down-shifts of -2 cm-1 in the positions

both the G- and G+ SWNTs modes occur when they are filled with ClAlPc

molecules. When covered with ClAlPc molecules even greater down-shifts of

-3 cm-1 and -4 cm-1 are observed.

131

SW

NT

sS

WN

Tm

od

eA

dd

itio

nal

Fre

qu

ency

Fre

qu

ency

shif

tR

elat

ive

Rel

ati

veR

elat

ive

Op

tim

um

fill

ing

Inte

nse

mod

esm

od

ified

freq

uen

cym

od

essh

ift

(fill

ed)

(cov

ered

)in

ten

sity

inte

nsi

tyin

ten

sity

dia

met

er/

nm

of

mole

cule

inw

ith

:/

cm-1

/cm

-1/c

m-1

/cm

-1(u

nm

od

ified

)(fi

lled

)(c

over

ed)

regio

n/

cm-1

NiT

PP

2.3

154

60

-20.

060.

05

0.04

-156

61

10.

210.

15

0.2

3-

159

31

01.

001.

00

1.0

0160

1-1

10.

160.

11

0.20

ClA

lPc

1.8

-154

8153

9-7

-70.

060.1

80.0

9156

7-2

-30.

190.1

90.1

8159

4-2

-41.

001.0

01.0

0160

4-1

-30.

080.1

00.0

8

NiP

c1.8

1548

155

7155

4-1

-50.

090.0

40.0

7156

6-1

00.

140.

19

0.19

159

3-1

01.

001.

00

1.00

160

50

00.

040.

07

0.0

8

Table

6.2:1.59eV

Excitation.Note:th

eintensitiesare

relativeto

theintensity

ofth

eG

+peak.

132

The relative intensities of the G- modes to the G+ modes are the same

for the unmodified and filled samples, however, a small decrease is observed

in the covered sample. The covered spectrum does not reflect this decrease

due to the presence of an intense mode located at 1539 cm-1. This mode is

not present in the unfilled SWNT spectrum and is therefore is most likely

due to a particularly strong Raman-active vibrational mode originating from

the ClAlPc molecule. Due to strong fluorescence it was not possible to reveal

a Raman spectrum from this molecule using this laser excitation, however,

the spectra acquired using the 2.34 eV and 3.83 eV lasers show an intense

molecular vibrational mode at ≈ 1520 cm-1 - it is likely that the 1539 cm-1

mode present in the filled and covered nanotube spectra is an upshift of the

1520 cm-1 mode which may have resulted from aggregation of the molecules.

It can be seen from Figure 6.20 (b) that the relative intensity of this peak is

greater for nanotubes covered with ClAlPc molecules.

For SWNTs modified with NiPc molecules, down-shifts of -1cm-1 are ob-

served in the positions of both the G- and G+ SWNT modes upon being

filled with NiPc molecules. No peak shifts are observed in the G- and G+

modes of SWNTs covered with NiPc molecules.

As well as the observed shifts in position, a increase of 5% in the relative

intensity of the G- peak is observed for SWNTs both filled and covered with

NiPc molecules.

In similarity with the ClAlPc modified SWNT spectra, there is a signifi-

cant well-defined peak located in the spectra of SWNTs modified with NiPc

molecules which is not present in the spectrum of the unmodified SWNTs.

This implies that this mode located at 1553 cm-1 originates from the NiPc

molecule. Indeed, this peak matches exactly the reported position of the

most intense Raman-active mode of the NiPc molecule [113]. In contrast to

what is observed for the ClAlPc molecule, this peak is larger in the spectrum

of SWNTs filled with NiPc molecules (see Figure 6.20 (c)).

133

Resonant Raman spectra resulting from a mixture of semiconduct-

ing and metallic SWNTs

Both the 2.34 eV and 3.83 eV laser excitation energies are such that they

are expected to resonate with the allowed transitions of both semiconducting

and metallic SWNTs. As a result, the resonant Raman spectra of the SWNT

G band acquired will be formed from the superposition of vibrational modes

of both electronic types.

As mentioned earlier in this chapter, the two most intense Raman-active

modes of the SWNT G band are the G- and G+ modes. Therefore, one

would expect the four most intense vibrational modes of a mixed G band

to be the G- metallic mode located at ≈ 1540 cm-1, the semiconducting G-

mode located at ≈ 1570 cm-1, the metallic G+ mode located at 1580 cm-1

and the semiconducting G+ mode located at ≈ 1590 cm-1. However, in

order to obtain a good fit to the experimental data it may be necessary to

include some of the weaker modes from the semiconducting G band. The

two strongest such modes are located at ≈ 1550 cm-1 and ≈ 1605 cm-1.

The SWNT G bands shown in the spectra of Figure 6.21 are fitted well

by the superposition of five Lorentzian line-shapes centred at approximately

1554 cm-1, 1574 cm-1, 1581 cm-1, 1598 cm-1 and 1608 cm-1.

The single broad mode centred at ≈ 1554 cm-1 possesses both a low

intensity and a low level of spectral detail. This mode could be attributed to

either a low intensity G band mode originating from semiconducting SWNTs

or the G- mode of metallic SWNTs, or a superposition of the two. An attempt

was made to fit a Breit Wigner Fano function to this peak - this was found to

fit poorly. This may indicate that the 1554 cm-1 mode has a predominately

semiconducting character, however, there is insufficient detail in the spectra

to identify the origin of this peak. Given the low level of spectral detail, it is

reasonable to put a very low weighting upon the changes in the position of

this mode.

To fit the 1577 cm-1 region of the G band, it was necessary to use two

Lorentizian line-shapes one centred at ≈ 1574 cm-1 and a second centred at

1580 cm-1.

134

The fitted peak located at 1574 cm-1 can be attributed to the G- mode

of semiconducting SWNTs. It possesses both the expected intensity relative

to the G+ semiconducting mode and is located at the expected position.

The expected position of the metallic G+ mode is 1580 cm-1, therefore it is

reasonable to assign the 1580cm-1 fitted peak to this mode. The amplitude

of this mode is approximately equal to that of the broad mode located at

1554 cm-1. This too is consistent with what would be expected in a metallic

SWNT spectrum, further supporting the argument for this mode’s metallic

character.

The superposition of these two modes forms a shoulder to the semicon-

ducting G+ mode. Changes in the shape of the shoulder will result from

either changes in the positions or the intensities of these modes. However,

due to the close proximity of these modes it is not possible to identify the

individual peaks this makes identifying their exact positions and intensities

very difficult. Unfortunately, this severely degrades the reliability of the fit-

ted positions of these peaks. As such, it is necessary to put a low weighting

to individual changes in the intensities and positions of the modes.

The very intense peak located at≈ 1594 cm-1 which dominates the G band

possesses both the expected position and relatively high intensity associated

with the semiconducting G+ mode. The relatively high intensity of this mode

makes identifying its position very easily and as such the position of the fitted

peak can be relied upon to be accurate. It is this peak which presents the

most reliable method for identifying charge transfer-induced peak shifts in

mixed SWNT samples.

The relatively low intensity fitted peak at ≈ 1608 cm-1 is so close to the

very intense G+ mode that it is not possible to observe the features of this

peak. Therefore, any changes in the position and or intensity of this peak

are unreliable.

In conclusion, due to the lack of spectral detail in the G band spectra

possessing Raman contributions from both types of nanotube, only the po-

sition of the dominant G+ semiconducting mode can be stated with a high

degree of confidence.

135

Therefore, changes in the position of G+ mode located at ≈ 1590 cm-1

will be used to inform upon whether charge transfer induced peak shifts have

occured.

Spectra acquired using the 2.34 eV laser

The SWNT G band spectra acquired using the 2.34 eV laser are shown in

Figure 6.21 - Table 6.3 summarises the changes observed in the modes of the

G band upon modification of the SWNTs with the organo-metallic molecules

of the study.

1400 1450 1500 1550 1600 1650

ELaser

= 2.34 eV Unmodified SWNTs SWNTs filled with NiTPP molecules SWNTs covered with NiTPP molecules

Inte

nsity / A

.U.

Raman shift / cm-1

1400 1450 1500 1550 1600 1650

Inte

nsity / A

.U.

Raman shift / cm-1

ELaser

=2.34 eV Unfilled SWNTs SWNTs covered with ClAlPc molecules

1400 1450 1500 1550 1600 1650

ELaser

=2.34 eV Unmodified SWNTs SWNTs filled with ClAlPc molecules

Inte

nsity / A

.U.

Raman shift / cm-1

1400 1450 1500 1550 1600 1650

ELaser

= 2.34 eV Unmodifed SWNTs SWNTs filled with NiPc molecules SWNTs covered with NiPc molecules

Inte

nsity / A

.U.

Raman shift / cm-1

ELaser=2.34 eV

Raman shift / cm-1 Raman shift / cm-1

Raman shift / cm-1 Raman shift / cm-1

(a) (b)

(c) (d)

Figure 6.21: (a) to (d) show resonant Raman spectra of SWNTs filled and covered withNiTPP, ClAlPc and NiPc molecules respectively, acquired using a 2.34 eV laser. Thespectra shown in (b) and (c) were acquired on different days and therefore could not beplotted on the same graph. The spectra of unmodifed SWNTs are shown in black as areference.

136

SW

NT

sS

WN

Tm

od

eA

dd

itio

nal

Fre

qu

ency

Fre

qu

ency

shif

tR

elat

ive

Rel

ati

veR

elat

ive

Op

tim

um

fill

ing

Inte

nse

mod

esm

od

ified

freq

uen

cym

od

essh

ift

(fill

ed)

(cov

ered

)in

ten

sity

inte

nsi

tyin

ten

sity

dia

met

er/

nm

of

mole

cule

inw

ith

:/

cm-1

/cm

-1/c

m-1

/cm

-1(u

nm

od

ified

)(fi

lled

)(c

over

ed)

regio

n/

cm-1

NiT

PP

2.3

155

4-3

60.

130.

13

0.14

157

40

-20.

220.

22

0.22

158

10

10.

280.

25

0.2

5159

80

01.

001.

00

1.0

0160

8-3

-10.

290.2

90.2

9

ClA

lPc

1.8

1520

155

03

-80.

200.

19

0.16

157

4-2

00.

440.

17

0.18

158

2-2

10.

120.

17

0.40

159

81

01.

001.

00

1.0

0160

46

90.

710.

14

0.1

1

NiP

c1.8

1545

154

2155

9-3

30.

170.

15

0.17

157

21

20.

320.

18

0.3

1158

10

10.

130.

18

0.1

7159

82

21.

001.

00

1.0

0161

13

10.

210.

14

0.2

4

Table

6.3:2.34eV

Excitation.Note:th

eintensitiesare

relativeto

theintensity

ofth

eG

+peak.

137

It can be seen from Figure 6.21 and Table 6.3 that there is no change in the

position of the semiconducting G+ mode upon modification of the SWNTs

with NiTPP molecules. However, there is also evidence of a decrease in the

intensity of the shoulder to the G+ mode upon both filling and covering.

The cluttered nature of the modes forming this shoulder make it difficult to

be sure which modes are responsible to the observed changes, however the

spectra imply that the reduction is greater for the filled nanotubes.

For SWNTs modified with ClAlPc molecules it can be seen that there is

a small up-shift of 1 cm-1 in the postion of semiconducting G+ mode when

the SWNTs were filled with ClAlPc molecules. However, no change in the

postion of this mode is observed when the SWNTs were covered. There

are also changes in relative intensities of the G- mode for both the SWNTs

fillied and covered with molecules, with the filling causing a reduction in the

intensity and the covering causing an enhancement. It is also worth noting

that there is no evidence of a molecule peak in either spectrum.

For SWNTs modified with NiPc molecules it can be seen that there is a

noticeable up-shift of 2 cm-1 in the postion of the semiconducting G+ mode

upon internal and external modification of the SWNTs with NiPc molecules.

There is also clear evidence of a peak at 1559 cm-1 which as before can be

attributed to the NiPc molecule. This peak possesses both the same position

and approximate intensity for both filled and covered SWNTs.

Spectra acquired using the 3.83 eV laser

The SWNT G band spectra acquired using the 3.83 eV laser are shown in

Figure 6.22 - Table 6.4 summarises the changes observed in the modes of the

G band upon modification of the SWNTs with the organo-metallic molecules

of the study.

138

1400 1450 1500 1550 1600 1650

ELaser= 3.83 eV Unmodified SWNTs SWNTs filled with NiTPP molecules SWNTs covered with NiTPP molecules

Inte

nsity

/ A.

U.

Raman shift / cm-1

(a)

1400 1450 1500 1550 1600 1650

ELaser= 3.83 eV Unmodified SWNTs SWNTs filled with ClAlPc molecules SWNTs covered with ClAlPc molecules

Inte

nsity

/ A.

U.

Raman shift / cm-1

(b)

1400 1450 1500 1550 1600 1650

ELaser= 3.83 eV Unmodified SWNTs SWNTs filled with NiPc molecules SWNTs covered with NiPc molecules

Inte

nsity

/ A.

U.

Raman shift / cm-1

(c)

Figure 6.22: (a) to (f) show resonant Raman spectra of SWNTs filled and covered withNiTPP, ClAlPc and NiPc molecules respectively acquired using a 3.83 eV laser. Thespectra of unmodifed SWNTs are shown in black as a reference.

139

SW

NT

sS

WN

Tm

od

eA

dd

itio

nal

Fre

qu

ency

Fre

qu

ency

shif

tR

elat

ive

Rel

ati

veR

elat

ive

Op

tim

um

fill

ing

Inte

nse

mod

esm

od

ified

freq

uen

cym

od

essh

ift

(fill

ed)

(cov

ered

)in

ten

sity

inte

nsi

tyin

ten

sity

dia

met

er/

nm

of

mole

cule

inw

ith

:/

cm-1

/cm

-1/c

m-1

/cm

-1(u

nm

od

ified

)(fi

lled

)(c

over

ed)

regio

n/

cm-1

NiT

PP

2.3

154

8-1

30.

090.

06

0.12

156

32

20.

270.

34

0.3

2157

36

70.

180.

09

0.2

4159

00

11.

001.

00

1.0

0

ClA

lPc

1.8

1522

155

0153

4(F

)0.

10156

215

38(C

)-2

30.

300.2

30.3

4157

3-2

0.29

0.18

159

1-2

-11.

001.0

01.0

0159

90.

31

NiP

c155

2(F

)1.8

154

815

53(C

)-3

0.08

0.06

156

04

0.11

0.1

5157

1-2

40.

430.

24

0.25

159

10

21.

001.

00

1.0

0160

20.

15

Table

6.4:3.83eV

Excitation.Note:th

eintensitiesare

relativeto

theintensity

ofth

eG

+peak.

140

It can be seen from Figure 6.22 and Table 6.4 that no change in the

position of the G+ SWNT peak located at 1590 cm-1 occurs upon filling with

NiTPP molecules, however a small up-shift of 1 cm-1 is observed for SWNTs

covered with NiTPP molecules. There may be some change in the relative

intensity of the shoulder to the G+ mode located at ≈ 1580 cm-1, but it is

difficult to be sure of the origins.

For SWNTs modified with ClAlPc molecules it can be seen that the po-

sition of the G+ semiconducting SWNT mode is down-shifted by -2 cm-1

and -1 cm-1 upon filling and covering with ClAlPc molecules respectively.

The fitted peaks imply a decrease in the intensity of the shoulder to the G+

peak for the covered SWNTs. However, the presence of the relatively intense

molecule peak in the spectrum make it appear that the shoulder has been

enhanced when is has not.

Both the filled and covered spectra show peaks which are not present

in the unmodified SWNT spectra at 1534 and 1538 cm-1 respectively. The

peaks are attributed to particularly intense vibrational mode originating from

the ClAlPc molecule as before. The intensity of this peak is greatest in the

spectrum of SWNTs filled with ClAlPc molecules.

The fitted peaks imply a decrease in the intensity of the shoulder to the

G+ peak for the filled SWNTs. However, the presence of the relatively intense

molecule peak in the spectrum make it appear that the shoulder has been

enhanced when is has not.

For SWNTs modified with NiPc molecules it can be seen that the G+

semiconducting mode is seen to be up-shifted by 2 cm-1 upon covering with

NiPc molecules, however no shift is observed for the SWNTs filled with NiPc

molecules.

The molecule peaks in both the filled and covered spectra are very intense,

in similarity with what was observed in the ClAlPc modified SWNTs the

molecule peak is greater for SWNTs filled with NiPc molecules. There is

also a small discrepancy in the positions of this peak, with it being located

at 1553 cm-1 in the covered spectrum and slightly lower at 1552 cm-1 in the

filled. This too is the same pattern as observed with the ClAlPc molecule.

141

The dominant nature of the molecule peak makes even assigning changes

in the relative intensities of the shoulders to the G+ mode difficult.

6.4.2 The SWNT RBM band

Spectra acquired using the 1.59 eV laser

The radial breathing mode spectra of the unmodified SWNTs acquired using

the 1.59 eV laser show four intense and sharp peaks and the spectra are fitted

well by six Lorentzian line shapes. In the low frequency region of the spectra,

depending upon the individual spectrum, one or two broad peaks are required

at approximately 110 cm-1 and 130 cm-1 - these fitted peaks can be assigned to

low frequency SWNT RBMs. Using equation 6.10 to calculate the diameters

of the SWNTs associated with these modes it can be calculated that they

originate from SWNTs of 2.34 nm and 1.96 nm in diameter respectively.

Such diameters would put these SWNTs close to the peak of the nanotube

diameter distribution function, meaning that there should be a large number

of these nanotubes. However, as previously mentioned the intensity of RBMs

originating from d > 2.0 nm are very low. This is likely to be the cause of

the very low intensity of these modes.

Given the breadth and relatively low intensity of these modes and the fact

that it is not possible to clearly identify the peak positions from the spectra, it

is reasonable to assign a low weighting to changes in their positions. However,

changes in the intensities of these modes may still be useful.

The remaining RBM peaks located at approximately 160 cm-1, 210 cm-1,

234 cm-1 and 268 cm-1 correspond to nanotubes of 1.6 nm, 1.2 nm, 1.0 nm

and 0.9 nm in diameter respectively. Using these calculated diameters in

combination with the Kataura plot shown in Figure 6.10 it is possible to

identify the electronic type of each of these SWNTs. From the Kataura plot

it can be seen that the RBM located at 160 cm-1 originates from metallic

SWNTs of ≈ 1.6 nm in diameter. The remaining RBM’s originate from

semiconducting SWNTs of ≈ 1.0 nm in diameter. All but the 234 cm-1 mode

are relatively sharp and intense.

142

Therefore, it is reasonable to view changes in the postion and intensity of

these modes with a high degree of confidence. It is changes in the postion and

intensity of these three modes which will be used to inform upon molecule-

induced changes to the SWNTs mechanical and electronic properties.

SWNTs modified with NiTPP molecules

It can be seen from Table 6.5 and from Figure 6.23 (a) and (b) that there

is no shift in the position of three most intense RBMs upon filling with

NiTPP molecules. However, there is an enhancement in the intensity in all

of the RBMs including the low intensity 114 cm-1 mode. In addition to these

increases in intensity, there is a change in peak dominance upon filling. In

the unmodified SWNT spectra, the two most intense peaks located at 164

cm-1 and 269 cm-1 possess approximately equal intensities, however, upon

filling with NiTPP molecules the 164 cm-1 peak becomes dominant.

SWNT modes / cm-1 ∆ω (Filling) / cm-1 ∆ω (Covering) / cm-1

114 9 9164 0 1211 0 1234 1 1269 0 0

Table 6.5: Vibrational modes of SWNTs modified with NiTPP molecules acquired usinga 1.59 eV laser.

143

100 150 200 250 3001000

2000

3000

4000

5000

6000

Unmodified SWNT's SWNT's filled with NiTPP

Inte

nsity

/ co

unts

per

sec

ond

Raman shift / cm-1

ELaser

= 1.59 eV

(a)

100 150 200 250 3001000

2000

3000

4000

5000

6000

Unmodified SWNT's SWNT's covered with NiTPP

Inte

nsity

/ co

unts

per

sec

ond

Raman shift / cm-1

ELaser

= 1.59 eV

(b)

100 150 200 250 300

1000

2000

3000

4000

5000

6000

Unmodified SWNT's SWNT's filled with ClAlc

Inte

nsity

/ co

unts

per

sec

ond

Raman shift / cm-1

ELaser

= 1.59 eV

(c)

100 150 200 250 300

1000

2000

3000

4000

5000

6000

7000

Unmodified SWNT's SWNT's covered with ClAlPc

Inte

nsity

/ co

unts

per

sec

ond

Raman shift / cm-1

ELaser

= 1.59 eV

(d)

100 150 200 250 300

400

500

600

700

800

900

Unmodified SWNT's SWNT's filled with NiPc

Inte

nsity

/ co

unts

per

sec

ond

Raman shift / cm-1

ELaser

= 1.59 eV

(e)

100 150 200 250 300

400

500

600

700

800

900

ELaser

= 1.59 eV Unmodified SWNT's SWNT's covered with NiPc

Inte

nsity

/ co

unts

per

sec

ond

Raman shift / cm-1

(f)

Figure 6.23: (a) to (f) show resonant Raman spectra of SWNTs filled and covered withNiTPP, ClAlPc and NiPc molecules respectively, acquired using a 1.59 eV laser. Thespectra of unmodifed SWNTs are shown in black as a reference.

In contrast, a small up-shift of 1 cm-1 in the positions of the first two

sharp peaks located at 164 cm-1 and 211 cm-1 was observed upon covering.

No up-shift is observed in the 264 cm-1 peak. In similarity to what is observed

in the NiTPP filled SWNT spectra, an increase in all of the sharp RBM’s was

observed when the SWNTs were covered with NiTPP molecules. However,

in contrast the peak dominance is reversed. In the NiTPP covered SWNT

spectra the 264 cm-1 peak is clearly dominant. In addition, no increase in

the intensity of the 110 cm-1 mode is observed.

144

In conclusion, the RBM’s of the SWNTs modified by both methods

show enhancements in the intensities of the modes after modification with

NiTPP molecules. However, the lower frequency modes originating from

larger diameter nanotubes are dominant when the SWNTs are filled with

NiTPP molecules, and the opposite is true for SWNTs covered with NiTPP

molecules.

SWNTs modified with ClAlPc molecules

It can be seen from Table 6.6 and Figure 6.23 (c) and (d) that an up-shift of

1 cm-1 is seen in the position of the 163 cm-1 SWNT RBM mode when the

SWNTs are filled with ClAlPc molecules. In addition to this, down-shifts of

2 cm-1 in the positions of the modes at 210 cm-1 and 268 cm-1 are observed.

As well as these changes in peak position there are a number of changes in the

intensity of the modes. The most striking changes in intensity originate from

the reversal of the dominance in the spectra. In the un-modified spectrum

the mode at 163 cm-1 is clearly dominant, however, upon filling, the mode at

268 cm-1 becomes dominant.

SWNT modes / cm-1 ∆ω (Filling) / cm-1 ∆ω (Covering) / cm-1

104 -7131 1163 1210 -2234 1268 -2

Table 6.6: Vibrational modes of SWNTs modified with ClAlPc molecules acquired usinga 1.59 eV laser.

Due to the high gradient of the spectrum acquired from the SWNTs

covered with ClAlPc molecules it was impossible to obtain a good fit to the

data. However, it is still possible to make a number of observations from the

relative intensities of the modes shown in the spectra. Firstly, there is an

increase in the intensity of the broad RBM located at approximately

131 cm-1. In addition to this there is a reversal in the dominance of the

RBM’s, with the 268 cm-1 mode becoming dominant after covering.

145

In conclusion, an up-shift of 1 cm-1 in the position of the 163 cm-1 RBM

originating from metallic SWNTs is observed upon filling of the SWNTs with

ClAlPc molecules. In contrast, the modes located at 210 cm-1 and 268 cm-1

originating from semiconducting SWNTs show down-shifts of 2 cm-1. It can

be seen that SWNTs modified with ClAlPc molecules using either method,

show both an enhancement of the broad peak located at 131 cm-1 originating

from semiconducting SWNTs of ≈ 2.0 nm in diameter and a shift in mode

dominance with the 268 cm-1 RBM becoming the most intense mode upon

modification.

SWNTs modified with NiPc molecules

The intensity of the spectra acquired from the NiPc samples is quite low;

this makes peak fitting more difficult and hence reduces the confidence of

the fitted peak positions of the RBM’s. Therefore, a greater weighting is

given to changes in the intensities of the RBM’s of these spectra and the

apparent changes in the positions of the fitted peaks will be ignored (table

of peak positions not included).

The unmodified SWNT spectra shown in Figure 6.23 (e) and (f) show four

relatively sharp peaks located at approximately 160 cm-1, 208 cm-1, 232 cm-1

and 266 cm-1. The positions of these modes before and after modification co-

incide very well, this implies that there is little if any change in position. The

most striking changes in the spectra upon modification of the SWNTs with

NiPc molecules is the reduction in intensity of the RBMs located at 160 cm-1

and 266 cm-1. The data imply a reduction in intensity because the intensities

of the 208 cm-1 and 232 cm-1 modes remain constant upon modification, but

the intensities of the other two modes decrease. The reduction in intensity

of the 160 cm-1 mode seems to be approximately equal for both forms of

modification, however the reduction in intensity of the 266 cm-1 mode seems

to be greater for SWNTs filled with NiPc molecules as opposed to covered.

146

In conclusion, the most striking differences between the unmodified SWNT

spectra and those modified with NiPc molecules come from decreases in the

intensities of the RBM’s located at 160 cm-1 and 266 cm-1 originating from

metallic SWNTs of 1.9 nm in diameter and semiconducting SWNTs of 0.9 nm

in diameter respectively. The decrease in the intensity of the 160 cm-1 mode

is approximately the same for both forms of modification. In contrast, the

intensity of the 266 cm-1 mode experiences the greater reduction in intensity

when the SWNTs are filled with molecules.

Conclusions

In conclusion, it can be seen from the spectra that for SWNTs modified with

the Pc type molecules the RBM located at approx. 268 cm-1 originating

from semiconducting SWNTs of 0.9 nm in diameter is the dominant mode.

This is in contradiction to what is observed in the spectra of the unmodified

SWNTs where the unmodified spectra it is the 168 cm-1 mode originating

from metallic SWNTs of ≈ 1.6 nm in diameter is dominant. In contrast to

what is observed in SWNTs filled with Pc type molecules, the 168 cm-1 mode

remained the dominant RBM of the SWNTs filled with NiTPP molecules.

However, upon covering with NiTPP molecules the 268 cm-1 RBM becomes

dominant in similarity to what is observed in the Pc modified SWNTs

Spectra acquired using the 2.34 eV laser

The RBM spectra acquired using the 2.34 eV laser can be fitted by at least

four Lorentzian line-shapes. However, it can be seen from the Kataura plot

that at this excitation energy the SWNT transition bands are much more

tightly grouped. This makes identification of the individual RBM’s more

difficult. In general, the spectra are fitted well by the superposition of five

Lorentzian line-shapes centred at 115 cm-1, 153 cm-1, 186 cm-1, 234 cm-1 and

264 cm-1.

Using equation 6.10 in combination with the Kataura plot it can be de-

termined that the RBM located at ≈ 115 cm-1 can be attributed to either

metallic or semiconducting SWNTs of 2.2 nm in diameter.

147

The broadness of the peak may be a reflection of the dense nature of the

allowed transitions at this energy, or as mentioned previously, may be a result

of the weak resonance of the RBM’s of SWNTs with d > 2.0 nm. Given the

breadth of this peak and that there is no clearly defined peak signature in

the spectra, it is reasonable to put a low weighting to changes in the position

of the peaks fitted to this mode, however, changes in the relative intensity of

this mode may be meaningful.

The other modes are considerably sharper than the 115 cm-1 mode and

hence the fitted peak positions are considerably more reliable. Using the

Kataura plot, the modes located at 153 cm-1 and 186 cm-1 can be seen to

originate from semiconducting SWNTs of 1.6 nm and 1.3 nm in diameter

respectively. The modes located at 234 cm-1 and 264 cm-1 can attributed to

metallic SWNTs of 1.0 nm and 0.9 nm in diameter respectively. The RBM

located at 234 cm-1 is very broad and of low intensity, therefore changes in

the position of this mode cannot be relied upon to be accurate.

In conclusion, the RBMs which are most reliable are those located at

163 cm-1 and 186 cm-1 originating from semiconducting SWNTs and the

mode located at 268 cm-1 originating from metallic SWNTs. Changes in the

positions and intensities of these modes will be utilized to identify changes

in the mechanical and electrical properties of the SWNTs upon doping with

molecules.

SWNTs modified with NiTPP molecules

It can be seen from Table 6.7 and Figure 6.24 (a) and (b) that the spectrum

of the unmodified SWNTs and those modified with NiTPP molecules are

very similar, however, there are a number of significant differences.

148

SWNT modes / cm-1 ∆ω (Filling) / cm-1 ∆ω (Covering) / cm-1

116 -2 -2152 7 6183 6 7265 8 6

Table 6.7: Vibrational modes of SWNTs modified with NiTPP molecules acquired usinga 2.34 eV laser.

50 100 150 200 250 3000

100

200

300

Unmodifed SWNT's SWNT's filled with NiTPP

Inte

nsity

/ co

unts

per

sec

ond

Raman shift / cm-1

ELaser

= 2.34 eV

(a)

50 100 150 200 250 3000

100

200

Unmodified SWNT's SWNT's covered with NiTPP

Inte

nsity

/ co

unts

per

sec

ond

Raman shift / cm-1

ELaser

= 2.34 eV

(b)

50 100 150 200 250 300

100

200

300

400

500

Unmodified SWNT's SWNT's filled with ClAlPc

Inte

nsity

/ co

unts

per

sec

ond

Raman shift / cm-1

ELaser

= 2.34 eV

(c)

50 100 150 200 250 300

100

200

300

400

500

Unmodified SWNT's SWNT's covered with ClAlPc

Inte

nsity

/ co

unts

per

sec

ond

Raman shift / cm-1

ELaser

= 2.34 eV

(d)

50 100 150 200 250 30050

100

150

200

250

300

350

400

450

Unmodified SWNT's SWNT's filled with NiPc

Inte

nsity

/ co

unts

per

sec

ond

Raman shift / cm-1

ELaser

= 2.34 eV

(e)

50 100 150 200 250 30050

100

150

200

250

300

350

400

Unmodified SWNT's SWNT's covered with NiPc

Inte

nsity

/ co

unts

per

sec

ond

Raman shift / cm-1

ELaser

= 2.34 eV

(f)

Figure 6.24: (a) to (f) show resonant Raman spectra of SWNTs filled and covered withNiTPP, ClAlPc and NiPc molecules respectively, acquired using a 2.34 eV laser. Thespectra of unmodified SWNTs are shown in black as a reference.

149

Upon filling with NiTPP molecules the 186 cm-1 SWNT RBM shows

a significant increase in intensity and becomes the dominant mode of the

spectrum. In contrast, the RBM located at 265 cm-1 shows a slight decrease

in intensity and is up-shifted by 8 cm-1 to 273 cm-1.

Upon covering with NiTPP molecules a small increase in the intensity of

the 186 cm-1 RBM, as well as a significant decrease in the intensity of the

mode located at 265 cm-1 is observed - there is also evidence of an up-shift

in the position of this mode.

In conclusion, the 186 cm-1 and 265 cm-1 SWNT RBM’s both show no-

ticeable changes after the SWNTs were modified with NiTPP molecules. The

intensity of the 186 cm-1 mode is seen to be enhanced upon both filling and

covering, however the enhancement is significantly greater for the SWNTs

filled with NiTPP molecules rather than covered. A reduction in intensity

of the 265 cm-1 mode and an up-shift in its position is observed for both

forms of modification, however the up-shift is far greater for filling and the

reduction in intensity is greater for covering.

SWNTs modified with ClAlPc molecules

It can be seen from the spectra shown in Figure 6.24 (c) and (d) and Table 6.8

that there is little or no significant change in either the intensity or position of

the 153 cm-1 or 189 cm-1 RBM’s upon either filling or covering of the SWNTs

with ClAlPc molecules. There is also little or no significant change in the

position or intensity of the 264 cm-1 mode upon covering of the SWNTs.

However, the peak at 264 cm-1 in the ClAlPc filled SWNT spectrum is seen

to be down-shifted by approximately 6 cm-1 and to be enhanced in intensity

by a significant amount.

150

SWNT modes / cm-1 ∆ω (Filling) / cm-1 ∆ω (Covering) / cm-1

119 -3 -10153 2 -3189 -1 -6234 7264 -6 3

Table 6.8: Vibrational modes of SWNTs modified with ClAlPc molecules acquired usinga 2.34 eV laser.

SWNTs modified with NiPc molecules

It can be seen from the spectra shown in Figure 6.24 (e) and (f) and Table 6.9

that there is evidence of a small increase in the intensity of the 186 cm-1 mode

in the spectra of SWNTs both filled and covered with NiPc molecules. In

addition to this, there is evidence of up-shifts in the positions of the 153 cm-1

and 186 cm-1 modes upon filling. There is also evidence of an enhancement in

the intensity of the 268 cm-1 mode upon both filling and covering. However,

this enhancement is significantly greater in the NiPc-covered SWNT spectra.

SWNT modes / cm-1 ∆ω (Filling) / cm-1 ∆ω (Covering) / cm-1

112 -5153 5 5186 4 2235 18268 1 0

Table 6.9: Vibrational modes of SWNTs modified with NiPc molecules acquired using a2.34 eV laser.

Conclusions

In conclusion, it can be seen from the spectra that the intensity of the RBM

located at ≈ 153 cm-1 is enhanced when the SWNTs are both filled and

covered using NiTPP and NiPc molecules. This enhancement is greatest for

SWNTs which have been filled with the molecules in both cases, however the

SWNTs filled with NiTPP molecules show the greatest enhancement.

151

The SWNTs modified with ClAlPc molecules show no change in the po-

sition or intensity of this mode.

The intensity of the SWNT RBM located at ≈ 265 cm-1 shows either a

small increase or no change in intensity when covered with NiPc and ClAlPc

molecules respectively. In contrast, the RBM of SWNTs covered with NiTPP

molecules shows a decrease in the intensity of the 265 cm-1 mode.

Spectra acquired using the 3.83 eV laser

It was not possible to acquire any useful RBM spectra from the nanotube

samples using the 3.83 eV laser. This was due to intrinsic limitations of the

spectrometer used.

6.5 Discussion

The Raman spectra described above revealed the following:

(i) Shifts in the positions of the G- and G+ SWNT modes and in the

position of the molecule modes.

(ii) The shifts in position are larger for certain molecules.

(iii) Changes in the relative intensity of the G- mode.

(iv) For a given molecule and type of SWNT modification employed,

different spectral changes were observed depending on which laser was

used.

A summary of the main results is given in Table 2.11.

152

ELaser Molecule ∆ G+ ∆ G+ dtsc dtscG doptimum

/ cm-1 / cm-1 / nm / nm / nm(Filled) (Covered)

1.59 eV 2.2 2.5NiTPP 1 0 2.3

ClAlPc -2 -4 1.8

NiPc -1 0 1.8

2.34 eV 1.8 2.0NiTPP 0 0 2.3

ClAlPc 1 0 1.8

NiPc 2 2 1.8

3.83 eV - -NiTPP 0 1 2.3

ClAlPc -2 -1 1.8

NiPc 0 2 1.8

Table 6.10: shows the shifts in the position of the G+ modes of spectra acquired fromSWNTs both filled and covered with each molecule acquired with each laser. ELaser is theenergy of the lasers used, ∆ G+ is the Raman shift in the position of the G+ mode observedupon filling and covering. dtsc indicates the diameter of the semiconducting nanotubes inresonance with each laser. dtscG indicates the diameter of the semiconducting nanotubesin resonance with photon scattered by the nanotube G mode. doptimum is the optimumfilling diameter for each molecule.

Such effects can be caused by a number of phenomena. The most likely

to cause such changes are (i) strain induced in the nanotubes as a result

of charge transfer between nanotube and molecule and (ii) structural strain

induced in the nanotubes by the filling and covering.

153

6.5.1 (i) Charge transfer

As discussed in section 6.2.5, charge transfer to or from dopant molecules

to nanotubes is known to result in strain being induced in the nanotube

structure either due to a weakening of the C-C bonds (n-type doping) or a

strengthening of the bonds (p-type doping).

Charge transfer to nanotubes can occur in one of two ways, either elec-

trons will be transferred from occupied states in the dopants to unoccupied

states in the conduction band of the nanotube (n-type doping) or electrons

will be donated from the nanotube valence band to unoccupied states in

the dopant (p-type doping). It has been demonstrated that charge transfer

between the SWNTs and organic molecules is controlled by the ionisation

potential (IP) and / or the electron affinity (EA) of the guest molecule [114].

The ionisation potential of a molecule can be thought of as the minimum

energy required to remove an electron. If the molecule in question was in

the ground state the ionisation potential would be the difference in energy

between the highest occupied molecular orbital (HOMO) and the vacuum

level. In contrast, the electron affinity is the energy released when an elec-

tron attaches to a molecule. In the ground state, this would be the energy

difference between the vacuum level and lowest unoccupied molecular orbital

(LUMO).

When the HOMO of the dopant molecule is greater in energy than the

conduction band of the nanotube as is the case in Figure 6.25 (a) then elec-

trons are donated from the molecule to the nanotube, and the nanotube

becomes n-type doped. If however, the valence band of the nanotube is

higher than the LUMO of the molecule (Figure 6.25 (b)) then electrons will

be transferred from the nanotube to the molecule and the nanotube will

become p-type doped.

154

Mole

cu

lar

orb

ita

l en

erg

y (

eV

)

Vacuum level

Guest molecule (a) Guest molecule (b)SWNT

Conduction band

Valence band

LUMO

HOMO

LUMO

HOMO

EA

IP

EA IP

e-

e-

(a) (b)

Figure 6.25: The possible charge transfer mechanisms between SWNTs and guestmolecules. (a) charge transfer from molecule to nanotube (b) charge transfer from nan-otube to molecule.

The charge transfer direction will depend very much upon the electronic

structure of the guest organic molecule. For example when the organic chain

molecule SPEEK which contains aromatic rings was added to the exterior of

semiconducting SWNTs of 1.48 nm by Zhu et. al. [115] charge transfer from

the valence bands of the SWNT to the LUMO of the SPEEK molecule was

observed. Charge transfer is confirmed by a noticeable up-shift in the SWNT

G band upon modification of the SWNTs with the SPEEK molecule.

In contrast, when SWNTs were functionalized with smaller aromatic

molecules such as Amph-TTF and TDD-TTF [98] and aromatic amines [96]

down-shifts in the positions of the SWNT G bands consistent with charge

transfer from the guest molecules to the SWNTs were observed.

Figure 6.26 shows the molecular orbitals (MOs) of the three molecules

used in this study. The HOMO and LUMO levels of the ClAlPc molecule

were determined experimentally [116], while those of the NiPc and porphyrin

molecules were calculated using density functional theory (DFT) [117, 118].

The band gaps and MOs of the phthalocyanine molecules are roughly the

same. The NiTPP molecule however, has both a larger band gap and a

higher LUMO than the other molecules.

155

Figure 6.26: The HOMO and LUMO energy levels [116–118] and optical absorptionspectra of the molecules used in this study ((a) ClAlPc, (b) NiPc and (c) NiTPP).

Optical absorption spectra acquired from the molecules have been in-

cluded to gauge the validity of the values obtained from literature.

156

Making a comparison between the two, it can be seen that the values

agree within ≈0.2 eV each other.

The other factor which determines whether charge transfer will occur is

the relative positions and occupancies of the valence and conduction bands

of the SWNTs upon which the guest molecules are attached. It has been

found that the Fermi level of nanotubes of greater than 1 nm in diameter is

approximated well by the work function of graphene (-4.66 eV) [119]. It has

also been found that semiconducting SWNTs are p-doped in ambient condi-

tions [120]; this means that the top of the valence band of semiconducting

SWNTs is partially vacant and ready to receive electrons from the HOMO

of a suitable molecule.

Another important aspect to consider is the quantum mechanical nature

of nanotubes - specifically the relationship between the band gap of the

nanotube and its diameter. In nanotubes the band gap is inversely propor-

tional to diameter. The band gap will determine the positions of the valence

and conduction bands of the nanotube. The band gaps between the V1 and

C1 van Hove singularities of semiconducting nanotubes (i.e. the E11s tran-

sition), calculated using the allowed transitions shown in the Kataura plot,

of diameters in the range 1.0 to 2.5 nm (encompassing the likely diameter

distribution used in this study) are given in Table 6.11.

Nanotube diameter / nm Band gap / eV

1.0 0.801.5 0.552.0 0.402.5 0.35

Table 6.11: The diameters and associated band gaps of semiconducting SWNTs.

Figure 6.27 shows the likely charge transfer mechanism between the ClAlPc

molecule and semiconducting nanotubes possessing diameters expected to

present in the samples.

157

It can be seen from this figure that, on making the

assumption that the obtained values for the molecular orbitals of the ClAlPc

molecule and band structure of the p-doped SWNTs are true, then charge

transfer from the HOMO of the molecules to the vacant states in the valence

band of the nanotubes will be possible. However, the amount of charge trans-

ferred will be reduced as the band gap decreases with increasing nanotube

diameter. Given that the band gap of the ClAlPc and NiPc molecules is very

similar, this mechanism, if accurate, is likely true for both.

Another factor which may have an effect upon the charge transfer be-

tween the molecules and the nanotubes is the extent to which the molecules

are distorted due interaction with the nanotube. It has been found using

density functional theory calculations that when distorted out of planarity

the HOMO level of the NiTPP molecule increases and the LUMO decreases

with the amount of distortion [121]. As discussed in chapter 3, the molecules

attached to nanotubes are likely to have some level of curvature induced dis-

tortion, the extent of which will depend upon the diameter of the

nanotube to which they are attached. Such distortions to the molecules both

encapsulated and attached to the exterior of SWNTs could have an effect

upon the charge transfer between the molecules and SWNTs. For exam-

ple, a distortion-induced increase in the height of the HOMO of an attached

molecule could result in charge transfer which would be unfavorable in the

planar molecule.

However, the calculations of Maji et al. [121] show that the increases in

the energy of the HOMO are expected to be small (less than 0.1 eV) and

therefore probably would not greatly affect the charge transfer between the

molecules and the nanotubes.

158

Mo

lecu

lar

orb

ita

l e

ne

rgy (

eV

)Vacuum level

0.0

-1.0

-2.0

-3.0

-4.0

-5.0

-6.0

Conduction band

Valence bande-

Conduction band

Valence band

Conduction band

Valence band

LUMO

HOMO

ClAlPc molecule P-SWNT- d = 1.0 nm P-SWNT- d = 2.0 nm P-SWNT- d = 3.0 nm

Increasing SWNT diameter

Figure 6.27: Schematic illustration of electron transfer between the ClAlPc moleculeand intrinsically p-doped SWNTs of various diameters.

This mechanism would fit well with the observed down-shifts in the G-

and G+ peaks observed when SWNTs modified with both ClAlPc and NiPc

molecules were excited by a 1.59 eV laser. At this excitation energy semicon-

ducting nanotubes of 2.0 nm in diameter are in resonance. This is close to the

optimum filling diameter of 1.8 nm, therefore molecule-nanotube interaction

should be maximised. The up-shift in the G- and G+ peaks observed when

the ClAlPc modified SWNTs were excited by a 2.34 eV laser and when the

NiPc modified SWNTs were excited by both the 2.34 eV and 3.83 eV lasers

are inconsistent with the above charge transfer mechanism. This may result

from different nanotube diameters being in resonance at these energies.

The NiTPP molecule is different both structurally and electronically and

is larger than the Phthalocyanine molecules. It possesses a larger HOMO-

LUMO gap of 2.9 eV meaning that its LUMO is likely to be even higher

in energy that that of the Pc molecules. The HOMO is expected to be at

approximately the same level as for the Pc molecules. In terms of charge

transfer this would imply that using the mechanism above, electron trans-

fer from the molecule to the nanotube should be favourable. However, the

up-shifts observed in the Raman spectra of SWNTs modified with NiTPP

excited with the 1.59 and 3.83 eV lasers would imply charge transfer from

the nanotubes to the molecule. This would be inconsistent with the above

charge transfer mechanism.

159

The molecular orbitals of the TPP molecule shown schematically in Fig-

ure 6.26 were obtained by density functional theory (DFT) calculations [118]

and while this method is known to give good indications of molecular HOMO-

LUMO gaps, it is less accurate in giving the exact position of the energy

levels. This is for the TPP molecule and not the NiTPP molecule but, the

HOMO and LUMOs of the two molecules should belong to the π and π*

orbitals and hence be the same. If charge transfer is responsible for the ob-

served down-shifts then the electron affinity of the NiTPP molecule would

have to be a lot larger - such a scenario in shown in Figure 6.28. In this sce-

nario only the band structure of larger diameter SWNTs have been included.

This is because the molecules are not expected to fit into the narrower tubes.

Mole

cu

lar

orb

ita

l en

erg

y (

eV

)

Vacuum level0.0

-1.0

-2.0

-3.0

-4.0

-5.0

-6.0

e-

Conduction band

Valence band LUMO

HOMO

NiTPP moleculeP-SWNT- d = 2.0 nm

Figure 6.28: Schematic illustration of a possible scenario for charge transfer from a2.0 nm semiconducting SWNT to the NiTPP molecule.

While the charge transfer mechanism described above could explain the

down-shifts observed in some of the G+ modes in spectra of the SWNTs

filled and covered with the phthalocyanine molecules provided that their

HOMOs are high enough, it is inconsistent with the up-shifts in the G+

modes observed in others such as any of the spectra of SWNTs functionalized

with NiTPP molecules.

160

6.5.2 (ii) Structural strain

The other way in which the nanotubes can be strained is by structural de-

formation of the nanotube walls, for instance by the introduction of foreign

material into the internal cavity of the nanotube or by attachment of mate-

rial to the exterior. Such a mechanism has been employed previously [122] to

explain up-shifts observed in the G+ modes of SWNTs doped with rubidium

atoms. It is likely that the greatest level of structural deformation-based

strain in filled nanotubes would result from tubes where the diameter of the

nanotube filled is less than the optimum filling diameter of the encapsulated

molecule [102].

The above argument could explain the up-shifts observed in some of the

G+ modes of SWNTs modified with molecular systems in this study, sum-

marised in Table 6.10.

The semiconducting SWNTs in resonance with the 1.59 eV laser and with

the G scattered photon have a diameters of 2.2 nm and 2.5 nm respectively.

These diameters match well with the expected optimum filling diameter for

the NiTPP molecule of 2.3 nm. It is possible that close to doptimum the filling

causes structural strain to the nanotube. If so, the above mechanism would

explain the observed up-shift in the G mode.

The semiconducting nanotubes in resonance with the 2.34 eV laser and

with the G scattered photon have a diameters of 1.8 nm and 2.0 nm respec-

tively. These tubes are too narrow for the NiTPP molecules to enter face-on,

therefore, it is likely that the resulting filling yield is either a very low (un-

detectable), this would be consistent with the lack of any peak shift in the

G+ mode.

The diameters of the nanotubes in resonance with the 2.34 eV laser are

very close to the optimum diameter for face-on filling of the NiPc molecules

as discussed previously for the NiTPP molecules filling the nanotubes probed

with the 1.59 eV laser, the up-shift in the G+ modes of the Pc-filled nanotubes

could be a result of filling-induced structural strain.

161

While individually the mechanisms do not entirely explain the observed

shifts in the G+ modes, a combination of the two mechanisms might serve to

explain the majority of the shifts. Not all of the shifts can be accounted for

by these two methods, for example the case of the up-shift observed in the

G band of the SWNTs covered with NiPc molecules acquired with the

3.83 eV laser where no up-shift is observed in the filled sample. It is possible

that this is due to some unknown effect - further study is needed to fully

explain such changes.

162

Chapter 7

Conclusions and future work

7.1 Conclusions

Endohedral functionalization via supercritical CO2 was undertaken in order

to produce encapsulation of compounds that are difficult to encapsulate oth-

erwise due to either their larger size or extreme air sensitivity. For this, two

experimental supercritical CO2 set-ups were developed, one for standard, sta-

ble encapsulants, and the other to enable the anaerobic encapsulation of air-

sensitive systems. Though equipment related technical difficulties prevented

the demonstration of encapsulation of air sensitive molecules (which would

have been the first of its kind were it to have been achieved) encapsulation

of planar molecules of large size (≈ 2 nm) has been achieved. These latter

systems are not suitable for thermally induced, diffusion-based encapsulation

due to their large size. Confinement in nanotubes of optimum diameter pro-

moted ordering of NiTPP molecules in row-like, self-assembled structures;

while disordered molecular arrangement dominated in larger diameter sys-

tems. High yield of molecular filling was also obtained for diameters larger

than an optimum value (of about 2.3 nm), though filling within diameters less

than optimum was also produced at low yield, and involved strong structural

strain to the molecule body.

163

The comparative endo- and exohedral systems (NiPc, ClAlPc and NiTPP)

were chosen so as to induce different degrees of electronic changes in the nan-

otube hosts, which were found to be mostly consistent with a combination

of charge transfer (controlled by the alignment of the molecular HOMO-

LUMO band gap to the energy spectrum of carbon nanotubes) and struc-

tural strain. These changes affect both the RBM and G modes of carbon

nanotubes. NiTPP, ClAlPc and NiPc molecules provided a set of systems

differing by only one specific parameter (e.g. central ion or body type, or size

of the HOMO-LUMO gap). Though the large NiTPP molecules and large

diameter Nanocyl SWNTs are perfectly matched in terms of geometry, the

decrease in size of the nanotubes’ electronic band gap that occurs in wide

Nanocyl nanotubes puts a limit on what can be achieved in terms of doping.

Finally, exohedral functionalization showed some degree of perturbation of

the electonic structure of the nanotubes demonstrating that the molecules

did attach to the outer surface of the nanotube despite perturbation by the

central metal ion of their aromatic system (which is supposed to promote π

stacking).

7.2 Future work

Future work could focus on the following directions:

(i) Comparative Raman/ IR studies where Raman mainly probes the

nanotube sub-system, IR spectroscopy would probe the molecular sub-system.

(ii) ClAlPc molecules possess an electronic dipole, this might effect what

ordering could be obtained inside of carbon nanotube templates, hence par-

allel HRTEM studies could reveal differences compared to their nickel-based

counterparts.

If encapsulated in an ordered form inside of a SWNT, the ClAlPc molecules

with their electronic dipole core could induce a modulated periodic potential

within the nanotube (see Figure 7.1). This could result in negative differen-

tial resistance in electronic transport or scanning tunnelling spectroscopy.

164

Figure 7.1: Electrostatic potential calculated along nanotube for a permanent dipoleperpendicular to the nanotube’s axis [123].

(iii) Another spectroscopic technique which could be employed is tip-

enhanced Raman spectroscopy (TERS). In a typical TERS experiment, a

laser is focused on the end of an AFM cantilever coated with gold - the tip

of the cantilever acts as a nanostructure to produce Raman signal enhance-

ment on a sample surface once the tip has been brought close enough. The

resolution of this process is ≈ 20 nm [124], allowing Raman spectra to be

acquired from very small areas [124,125], hence the local effect of individual

or a small number of encapsulates could be probed. TERS has been used

successfully to acquire Raman spectra from carbon nanotubes [124,125]. Use

of this technique on the nanotube hybrids of this study, if sufficiently dis-

persed, would allow for the nanotubes to be probed in their un-bundled state

and would allow for spectra to be acquired from individual semiconducting

and metallic SWNTs. This would make assessing shifts in the peak positions

of the G- modes much easier.

(iv) Finally, the hybrids produced in this work should possess paramag-

netism. By studying their magnetic properties, new physics may be discov-

ered.

165

Bibliography

[1] S. Iijima et al. Helical microtubules of graphitic carbon. nature,

354(6348):56–58, 1991.

[2] A. K. Geim and K. S. Novoselov. The rise of graphene. Nature mate-

rials, 6(3):183–191, 2007.

[3] M. S. Dresselhaus R. Saito, G. Dresselhaus. Physical Properties of

Carbon Nanotubes. Imperial College Press, 2007.

[4] D. Tasis, N. Tagmatarchis, A. Bianco, and M. Prato. Chemistry of

carbon nanotubes. Chemical Reviews, 106(3):1105–1136, 2006.

[5] D. A. Britz and A. N. Khlobystov. Noncovalent interactions of

molecules with single walled carbon nanotubes. Chem. Soc. Rev.,

35:637–659, 2006.

[6] R. R. Meyer, J. Sloan, R. E. Dunin-Borkowski, A. I. Kirkland, M. C.

Novotny, S. R. Bailey, J. L. Hutchison, and M. L. H. Green. Discrete

atom imaging of one-dimensional crystals formed within single-walled

carbon nanotubes. Science, 289(5483):1324–1326, 2000.

[7] D. A. Britz, A. N. Khlobystov, J. Wang, A. S. O’Neil, M. Poliakoff,

A. Ardavan, and G. A. D. Briggs. Selective host–guest interaction of

single-walled carbon nanotubes with functionalised fullerenes. Chemi-

cal communications, (2):176–177, 2004.

[8] A. Ilie, J. S. Bendall, K. Nagaoka, S. Egger, T. Nakayama, and

S. Crampin. Encapsulated inorganic nanostructures: A route to sizable

modulated, noncovalent, on-tube potentials in carbon nanotubes. ACS

nano, 5(4):2559–2569, 2011.

166

[9] M. Monthioux, E. Flahaut, and J. P. Cleuziou. Hybrid carbon nan-

otubes: Strategy, progress, and perspectives. Journal of materials re-

search, 21(11):2774–2793, 2006.

[10] W. Zhiyong, S. Zujin, and G. Zhennan. New phenomena of materials

confined in nano space. PROGRESS IN CHEMISTRY, 21(11):2435–

2444, 2009.

[11] P. M. F. J. Costa, D. Golberg, M. Mitome, S. Hampel, A. Leonhardt,

B. Buchner, and Y. Bando. Stepwise current-driven release of attogram

quantities of copper iodide encapsulated in carbon nanotubes. Nano

letters, 8(10):3120–3125, 2008.

[12] S. Brahim, S. Colbern, R. Gump, and L. Grigorian. Tailoring gas

sensing properties of carbon nanotubes. Journal of Applied Physics,

104(2):024502–024502, 2008.

[13] J. H. Warner, A. A. R. Watt, L. Ge, K. Porfyrakis, T. Akachi, H. Oki-

moto, Y. Ito, A. Ardavan, B. Montanari, J. H. Jefferson, et al. Dy-

namics of paramagnetic metallofullerenes in carbon nanotube peapods.

Nano letters, 8(4):1005–1010, 2008.

[14] S. Y. Hong, G. Tobias, K. T. Al-Jamal, B. Ballesteros, H. Ali-Boucetta,

S. Lozano-Perez, P. D. Nellist, R. B. Sim, C. Finucane, S. J. Mather,

et al. Filled and glycosylated carbon nanotubes for in vivo radioemitter

localization and imaging. Nature materials, 9(6):485–490, 2010.

[15] A. N. Khlobystov, D. A. Britz, O’neil S. A. Martyn P. Wang, J., and

G.A.D. Briggs. Low temperature assembly of fullerene arrays in single-

walled carbon nanotubes using supercritical fluids. Journal of Materials

Chemistry, 14(19):2852–2857, 2004.

[16] L. Li, A. N. Khlobystov, J. G. Wiltshire, G. A. D. Briggs, and R. J.

Nicholas. Diameter-selective encapsulation of metallocenes in single-

walled carbon nanotubes. Nature materials, 4(6):481–485, 2005.

[17] P. Peumans and S. R. Forrest. Very-high-efficiency double-

heterostructure copper phthalocyanine/c-60 photovoltaic cells. Applied

Physics Letters, 79(1):126–128, 2001.

167

[18] Y. Qiu, Y. Gao, P. Wei, and L. Wang. Organic light-emitting

diodes with improved hole-electron balance by using copper phthalo-

cyanine/aromatic diamine multiple quantum wells. Applied physics

letters, 80(15):2628–2630, 2002.

[19] D. R. Kauffman, O. Kuzmych, and A. Star. Interactions between single-

walled carbon nanotubes and tetraphenyl metalloporphyrins: Corre-

lation between spectroscopic and fet measurements. The Journal of

Physical Chemistry C, 111(9):3539–3543, 2007.

[20] J. De Paula P. Atkins. Physical Chemistry. Oxford University Press,

2006.

[21] A. Ilie. Nanoscience lecture notes.

[22] N. D. Mermin N. W. Ashcroft. Solid State Physics. Brooks/Cole, 1976.

[23] J. Maultzsch S. Reich, C. Thomsen. Carbon Nanotubes Basic Concepts

and Physical Properties. Wiley-VCH, 2004.

[24] A. L. Kalamkarov, A.V. Georgiades, S.K. Rokkam, VP Veedu, and

M.N. Ghasemi-Nejhad. Analytical and numerical techniques to pre-

dict carbon nanotubes properties. International Journal of Solids and

Structures, 43(22):6832–6854, 2006.

[25] Nanodic.com, 2008-2011. armchair nanotube [online]. available from:

http://www.nanodic.com/picturepreview.php?ch=carbon&term=armchair

nanotube&g=4.[accessed november 2013].

[26] J.W. Mintmire and C.T. White. Universal density of states for carbon

nanotubes. Physical Review Letters, 81(12):2506, 1998.

[27] M.S. Dresselhaus, G. Dresselhaus, and A. Jorio. Unusual properties

and structure of carbon nanotubes. Annu. Rev. Mater. Res., 34:247–

278, 2004.

[28] M. S. Dresselhaus, G. Dresselhaus, R. Saito, and A. Jorio. Raman spec-

troscopy of carbon nanotubes. Physics Reports, 409(2):47–99, 2005.

168

[29] S. Niyogi, M. A. Hamon, H. Hu, B. Zhao, P. Bhowmik, R. Sen, M. E.

Itkis, and R. C. Haddon. Chemistry of single-walled carbon nanotubes.

Accounts of Chemical Research, 35(12):1105–1113, 2002.

[30] Z. Chen, W. Thiel, and A. Hirsch. Reactivity of the convex and con-

cave surfaces of single-walled carbon nanotubes (swcnts) towards ad-

dition reactions: Dependence on the carbon-atom pyramidalization.

ChemPhysChem, 4(1):93–97, 2003.

[31] L.A. Girifalco, M. Hodak, and R. S. Lee. Carbon nanotubes, bucky-

balls, ropes, and a universal graphitic potential. Physical Review B,

62(19):13104, 2000.

[32] L.A. Girifalco and M. Hodak. Van der waals binding energies in

graphitic structures. Physical Review B, 65(12):125404, 2002.

[33] G. Stan, M. J. Bojan, S. Curtarolo, S. M. Gatica, and M. W. Cole.

Uptake of gases in bundles of carbon nanotubes. Physical Review B,

62(3):2173, 2000.

[34] J. Wang, M. K. Kuimova, M. Poliakoff, G. A. D. Briggs, and An-

drei N Khlobystov. Encapsulation and ir probing of cube-shaped oc-

tasilasesquioxane h8si8o12 in carbon nanotubes. Angewandte Chemie

International Edition, 45(31):5188–5191, 2006.

[35] K. Muthukumar, H. O. Jeschke, R. Valentı, E. Begun, J. Schwenk,

F. Porrati, and M. Huth. Spontaneous dissociation of co2 (co) 8 and

autocatalytic growth of co on sio2: A combined experimental and the-

oretical investigation. Beilstein Journal of Nanotechnology, 3(1):546–

555, 2012.

[36] K. Schulte, C. Yan, M. Ahola-Tuomi, A . Strozecka, P.J. Moriarty, and

A.N. Khlobystov. Encapsulation of cobalt phthalocyanine molecules in

carbon nanotubes. In Journal of Physics: Conference Series, volume

100, page 012017, 2008.

[37] Computer model of the nitpp molecule produced by dr. adelina ilie

using the software crystalmaker.

169

[38] Computer model of the nitpp molecule inside a (25,0) swcnt of 2nm

diameter produced by dr. adelina ilie using the software crystalmaker.

[39] H. Li, B. Zhou, Y. Lin, L. Gu, W. Wang, K.A. Shiral Fernando, S. Ku-

mar, L. F. Allard, and Y. Sun. Selective interactions of porphyrins

with semiconducting single-walled carbon nanotubes. Journal of the

American Chemical Society, 126(4):1014–1015, 2004.

[40] X. Wang, Y. Liu, W. Qiu, and D. Zhu. Immobilization of tetra-tert-

butylphthalocyanines on carbon nanotubes: a first step towards the

development of new nanomaterials. Journal of Materials Chemistry,

12(6):1636–1639, 2002.

[41] F. Tournus, S. Latil, M. I. Heggie, and J.-C. Charlier. π-stacking inter-

action between carbon nanotubes and organic molecules. Phys. Rev.

B, 72:075431, 2005.

[42] S. Cambre, W. Wenseleers, E. Goovaerts, and D. E. Resasco. Determi-

nation of the metallic/semiconducting ratio in bulk single-wall carbon

nanotube samples by cobalt porphyrin probe electron paramagnetic

resonance spectroscopy. ACS nano, 4(11):6717–6724, 2010.

[43] D.J. Shu and X.G. Gong. Curvature effect on surface diffusion: The

nanotube. The Journal of Chemical Physics, 114:10922, 2001.

[44] H. Kataura, Y. Maniwa, M. Abe, A. Fujiwara, T. Kodama, K. Kikuchi,

H. Imahori, Y. Misaki, S. Suzuki, and Y. Achiba. Optical properties of

fullerene and non-fullerene peapods. Applied Physics A, 74(3):349–354,

2002.

[45] H. Rauf T. Pichler J. Bernardi H. Peterlik L. Korecz F. Flp F. Simon,

H. Kuzmany and A. Jnossy. Low temperature fullerene encapsulation

in single wall carbon nanotubes: synthesis of n@c60@swcnt. Chemical

Physics Letters, 383:362–367, 2004.

[46] K. Suenaga T. Ichihashi A. Hashimoto M. Yudasaka, K. Ajima and

S. Iijima. Nano-extraction and nano-condensation for c60 incorporation

into single-wall carbon nanotubes in liquid phases. Chemical Physics

Letters, 380:42–46, 2003.

170

[47] Occupational safety and health guideline for cobalt carbonyl (u.s. de-

partment of health and human services).

[48] K. Smith. Porphyrins and Metalloporphyrins. Elsevier Scientific Pub-

lishing company, 1975.

[49] A. L. Thomas F. H. Moser. The Phthalocyanines Volume 1 Properties.

CRC Press, 1983.

[50] E. Dujardin, T. W. Ebbesen, H. Hiura, and K. Tanigaki. Capillarity

and wetting of carbon nanotubes. Science, 265(5180):pp. 1850–1852,

1994.

[51] T. W. Ebbesen. Wetting, filling and decorating carbon nanotubes.

Journal of Physics and Chemistry of Solids, 57:951–955, 1996.

[52] R. Malhotra R. S. Ruoff, Doris S. Tse and Donald C. Lorents. Solubility

of fullerene (c60) in a variety of solvents. The Journal of Physical

Chemistry, 97:3379–3383, 1993.

[53] A. Bondi. van der waals volumes and radii. The Journal of Physical

Chemistry, 68:441–451, 1964.

[54] N. A. Lange; J. A. Dean. LANGE’S HANDBOOK OF CHEMISTRY

14th Edition. McGraw-Hill, 1992.

[55] M. Roth, K. Maag, G. M. Schneider, and Dirk T. Solution properties

of c60 and c70 fullerenes and c21 to c40 n-alkanes at infinite dilution in

carbon dioxide from supercritical fluid chromatography. The Journal

of Physical Chemistry B, 105(42):10373–10378, 2001.

[56] S.A. Westwood. Supercritical Fluid extraction and its use on in chro-

matographic sampler preparation. Blackie academic and professional,

1992.

[57] S. Rizvi. Supercritical fluid processing of food and biomaterials. Chap-

man and Hall, 1994.

[58] J. M. DeSimone. Practical approaches to green solvents. Science,

297(5582):799–803, 2002.

171

[59] S. Iijima and T. Ichihashi. Single-shell carbon nanotubes of 1-nm di-

ameter. Nature, 363:603 – 605, 1993.

[60] NanocylTM nc1000 10 march 2009 - v03. Manufacturers data sheet

containing information on the NC1100 nanotubes used in this study.

[61] Y. Li, W. Kim, Y. Zhang, M. Rolandi, D. Wang, and H. Dai. Growth

of single-walled carbon nanotubes from discrete catalytic nanoparticles

of various sizes. The Journal of Physical Chemistry B, 105(46):11424–

11431, 2001.

[62] V. Georgakilas, D. Voulgaris, E. Vazquez, Maurizio P., D. M. Guldi,

A. Kukovecz, and H. Kuzmany. Purification of hipco carbon nanotubes

via organic functionalization. Journal of the American Chemical Soci-

ety, 124(48):14318–14319, 2002.

[63] Y. Feng, H. Zhang, Y. Hou, T. P. McNicholas, D. Yuan, S. Yang,

L. Ding, W. Feng, and Jie Liu. Room temperature purification of few-

walled carbon nanotubes with high yield. ACS nano, 2(8):1634–1638,

2008.

[64] K. Tohji, H. Takahashi, Y. Shinoda, N. Shimizu, B. Jeyadevan, I. Mat-

suoka, Y. Saito, A. Kasuya, S. Ito, and Y. Nishina. Purification proce-

dure for single-walled nanotubes. The Journal of Physical Chemistry

B, 101(11):1974–1978, 1997.

[65] P.M. Ajayan, T.W. Ebbesen, T. Ichihashi, S. Iijima, K.S. Tanigaki, and

H. Hiura. Opening carbon nanotubes with oxygen and implications for

filling. Nature, 362:522 – 525, 1993.

[66] K. Sathiyamoorthy, C. Vijayan, and S. Varma. Nonlinear optical re-

sponse of chloroaluminiumphthalocyanine encapsulated by silica core-

shell particles. Langmuir, 24(14):7485–7491, 2008.

[67] S. Wischnitzer. Introduction to Electron Microscopy. Pergamon Press,

1981.

[68] B. W. Smith and D. E. Luzzi. Electron irradiation effects in single wall

carbon nanotubes. Journal of applied Physics, 90(7):3509–3515, 2001.

172

[69] F. Banhart. Irradiation effects in carbon nanostructures. Reports on

Progress in Physics, 62(8):1181, 1999.

[70] A. Ilie, S. Crampin, L. Karlsson, and M. Wilson. Repair and stabiliza-

tion in confined nanoscale systemsinorganic nanowires within single-

walled carbon nanotubes. Nano Research, 5(12):833–844, 2012.

[71] Crystalmaker,2.0; crystalmaker software: Oxford, 2006.

[72] A. Gomez-Rodrıguez, L. M. Beltran-del Rıo, and R. Herrera-Becerra.

Simulatem: Multislice simulations for general objects. Ultramicroscopy,

110(2):95–104, 2010.

[73] A. Thust C. Kbel. ’TrueImage: Introduction to Focal-Series Recon-

struction’ in Nato Science Series E. Kluwer Academic Publishers, 2005.

[74] K. Schulte, J. C. Swarbrick, N. A. Smith, F. Bondino, E. Magnano,

and A. N. Khlobystov. Assembly of cobalt phthalocyanine stacks inside

carbon nanotubes. Advanced Materials, 19(20):3312–3316, 2007.

[75] S. Okada and H. Segawa. Substituent-control exciton in j-aggregates

of protonated water-insoluble porphyrins. Journal of the American

Chemical Society, 125(9):2792–2796, 2003.

[76] Y. Fujita, S. Bandow, and S. Iijima. Formation of small-diameter car-

bon nanotubes from ptcda arranged inside the single-wall carbon nan-

otubes. Chemical physics letters, 413(4):410–414, 2005.

[77] M. C. Tobin. Laser Raman spectroscopy. Wiley-Interscience, 1971.

[78] P. J. Hendra T. R. Gilson. Laser Raman Spectroscopy. Wiley-

Interscience, 1970.

[79] S. Sasic. Pharmaceutical applications of Raman spectroscopy. Wiley-

Interscience, 2008.

[80] R. Saito, M. Hofmann, G. Dresselhaus, A. Jorio, and M.S. Dresselhaus.

Raman spectroscopy of graphene and carbon nanotubes. Advances in

Physics, 60(3):413–550, 2011.

173

[81] A.M. Rao, E. Richter, S. Bandow, B. Chase, P.C. Eklund, K.A.

Williams, S. Fang, K.R. Subbaswamy, M. Menon, A. Thess, et al.

Diameter-selective raman scattering from vibrational modes in carbon

nanotubes. Science, 275(5297):187–191, 1997.

[82] M. J. OConnell, S. Sivaram, and S. K. Doorn. Near-infrared resonance

raman excitation profile studies of single-walled carbon nanotube in-

tertube interactions: A direct comparison of bundled and individually

dispersed hipco nanotubes. Physical Review B, 69(23):235415, 2004.

[83] S. Piscanec, M. Lazzeri, J. Robertson, A. C. Ferrari, and F. Mauri.

Optical phonons in carbon nanotubes: Kohn anomalies, peierls distor-

tions, and dynamic effects. Physical Review B, 75(3):035427, 2007.

[84] S.D.M. Brown, A. Jorio, P. Corio, M.S. Dresselhaus, G. Dresselhaus,

R. Saito, and K. Kneipp. Origin of the breit-wigner-fano lineshape of

the tangential g-band feature of metallic carbon nanotubes. Physical

Review B, 63(15):155414, 2001.

[85] S. K. Das, A. S.D. Sandanayaka, N. K. Subbaiyan, M. E. Zan-

dler, O. Ito, and F. D’Souza. Functionalization of diameter-sorted

semiconductive swcnts with photosensitizing porphyrins: Syntheses

and photoinduced electron transfer. Chemistry-A European Journal,

18(36):11388–11398, 2012.

[86] H. Zhao, Y. Zhu, C. Chen, L. He, and J. Zheng. Synthesis, charac-

terization, and photophysical properties of covalent-linked ferrocene–

porphyrin–single-walled carbon nanotube triad hybrid. Carbon,

50(13):4894–4902, 2012.

[87] B. Ballesteros, G. de la Torre, C. Ehli, G.M. Aminur Rahman,

F. Agullo-Rueda, D. M. Guldi, and T. Torres. Single-wall carbon nan-

otubes bearing covalently linked phthalocyanines-photoinduced elec-

tron transfer. Journal of the American Chemical Society, 129(16):5061–

5068, 2007.

[88] T. Mugadza and T. Nyokong. Covalent linking of ethylene

amine functionalized single-walled carbon nanotubes to cobalt (ii)

174

tetracarboxyl-phthalocyanines for use in electrocatalysis. Synthetic

Metals, 160(19):2089–2098, 2010.

[89] W. Chidawanyika and T. Nyokong. Characterization of amine-

functionalized single-walled carbon nanotube-low symmetry phthalo-

cyanine conjugates. Carbon, 48(10):2831–2838, 2010.

[90] C. Meyer, C. Besson, R. Frielinghaus, A. Saelhoff, H. Flototto,

L. Houben, P. Kogerler, and C. M. Schneider. Covalent functional-

ization of carbon nanotubes with tetramanganese complexes. physica

status solidi (b), 249(12):2412–2415, 2012.

[91] M.V. Kharlamova, L.V. Yashina, A.A. Eliseev, A.A. Volykhov, V.S.

Neudachina, M.M. Brzhezinskaya, T.S. Zyubina, A.V. Lukashin, and

Yu D Tretyakov. Single-walled carbon nanotubes filled with nickel

halogenides: Atomic structure and doping effect. physica status solidi

(b), 249(12):2328–2332, 2012.

[92] F.H. Monteiro, D.G. Larrude, M.E.H. Maia da Costa, and F.L. Freire.

Estimating the boron doping level on single wall carbon nanotubes

using raman spectroscopy. Materials Letters, 2012.

[93] S. Suzuki and H. Hibino. Characterization of doped single-wall carbon

nanotubes by raman spectroscopy. Carbon, 49(7):2264–2272, 2011.

[94] G. Jayamurugan, K.S. Vasu, Y.B.R.D. Rajesh, S. Kumar, V. Va-

sumathi, P.K. Maiti, A.K. Sood, and N. Jayaraman. Interaction of

single-walled carbon nanotubes with poly (propyl ether imine) den-

drimers. The Journal of Chemical Physics, 134:104507, 2011.

[95] L.G. Moura, L.M. Malard, M.A. Carneiro, P. Venezuela, R. B. Ca-

paz, D. Nishide, Y. Achiba, H. Shinohara, and M.A. Pimenta. Charge

transfer and screening effects in polyynes encapsulated inside single-

wall carbon nanotubes. Physical Review B, 80(16):161401, 2009.

[96] R. B. Koizhaiganova, D. H. Hwang, C. J. Lee, S. Roth, and U. Dettlaff-

Weglikowska. N-type doping effect of single-walled carbon nanotubes

with aromatic amines. physica status solidi (b), 247(11-12):2793–2796,

2010.

175

[97] M.V. Kharlamova and J.J. Niu. Comparison of metallic silver and cop-

per doping effects on single-walled carbon nanotubes. Applied Physics

A, 109(1):25–29, 2012.

[98] A. Ghosh, K. V. Rao, R. Voggu, and S. J. George. Non-covalent func-

tionalization, solubilization of graphene and single-walled carbon nan-

otubes with aromatic donor and acceptor molecules. Chemical Physics

Letters, 488(4):198–201, 2010.

[99] Dr. s. maruyama, 2002. kataura-plot for resonant raman

[online]. department of mechanical engineering, the univer-

sity of tokyo, tokyo. available from: http://www.photon.t.u-

tokyo.ac.jp/ maruyama/kataura/kataura.html [accessed november

20013].

[100] S. Reich, C. Thomsen, and P. Ordejon. Electronic band struc-

ture of isolated and bundled carbon nanotubes. Physical Review B,

65(15):155411, 2002.

[101] F. Hennrich, R. Krupke, S. Lebedkin, K. Arnold, R. Fischer, D. E.

Resasco, and M. M. Kappes. Raman spectroscopy of individual single-

walled carbon nanotubes from various sources. The Journal of Physical

Chemistry B, 109(21):10567–10573, 2005.

[102] A. Ilie, J.S. Bendall, D. Roy, E. Philp, and M.L.H. Green. Effects

of ki encapsulation in single-walled carbon nanotubes by raman and

optical absorption spectroscopy. The Journal of Physical Chemistry B,

110(28):13848–13857, 2006.

[103] M. Steiner, M. Freitag, J. C. Tsang, V. Perebeinos, A. A. Bol, A. V.

Failla, and P. Avouris. How does the substrate affect the raman and

excited state spectra of a carbon nanotube? Applied Physics A,

96(2):271–282, 2009.

[104] Y. Zhang, H. Son, J. Zhang, J. Kong, and Z. Liu. Laser-heating effect

on raman spectra of individual suspended single-walled carbon nan-

otubes. The Journal of Physical Chemistry C, 111(5):1988–1992, 2007.

176

[105] H.D. Li, K.T. Yue, Z.L. Lian, Y. Zhan, L.X. Zhou, S.L. Zhang, Z.J.

Shi, Z.N. Gu, B.B. Liu, R.S. Yang, et al. Temperature dependence of

the raman spectra of single-wall carbon nanotubes. Applied Physics

Letters, 76(15):2053–2055, 2000.

[106] Z. Zhou, X. Dou, L. Ci, L. Song, D. Liu, Y. Gao, J. Wang, L. Liu,

W. Zhou, S. Xie, et al. Temperature dependence of the raman spectra

of individual carbon nanotubes. The Journal of Physical Chemistry B,

110(3):1206–1209, 2006.

[107] A. Jorio, C. Fantini, M.S.S. Dantas, M.A. Pimenta, A.G. Souza Filho,

G. G. Samsonidze, V.W. Brar, G. Dresselhaus, M.S. Dresselhaus, A.K.

Swan, et al. Linewidth of the raman features of individual single-wall

carbon nanotubes. Physical Review B, 66(11):115411, 2002.

[108] N. R. Raravikar, P. Keblinski, A. M. Rao, M. S. Dresselhaus, L. S.

Schadler, and P. M. Ajayan. Temperature dependence of radial breath-

ing mode raman frequency of single-walled carbon nanotubes. Physical

Review B, 66(23):235424, 2002.

[109] C. Fantini, A. Jorio, M. Souza, M.S. Strano, M.S. Dresselhaus, and

M.A. Pimenta. Optical transition energies for carbon nanotubes from

resonant raman spectroscopy: Environment and temperature effects.

Physical review letters, 93(14):147406, 2004.

[110] M. Takase, H. Nabika, S. Hoshina, M. Nara, K. Komeda, R. Shito,

S. Yasuda, and K. Murakoshi. Local thermal elevation probing of metal

nanostructures during laser illumination utilizing surface-enhanced ra-

man scattering from a single-walled carbon nanotube. Phys. Chem.

Chem. Phys., 15(12):4270–4274, 2013.

[111] S. Chiashi, Y. Murakami, Y. Miyauchi, and S. Maruyama. Temperature

dependence of raman scattering from single-walled carbon nanotubes:

Undefined radial breathing mode peaks at high temperatures. Japanese

Journal of Applied Physics-Part 1 Regular Papers and Short Notes,

47(4):2010–2015, 2008.

[112] W. Wenseleers, S. Cambre, J. Culin, A. Bouwen, and E. Goovaerts. Ef-

fect of water filling on the electronic and vibrational resonances of car-

177

bon nanotubes: Characterizing tube opening by raman spectroscopy.

Advanced Materials, 19(17):2274–2278, 2007.

[113] Z. Liu, X. Zhang, Y. Zhang, and J. Jiang. Theoretical investigation of

the molecular, electronic structures and vibrational spectra of a series

of first transition metal phthalocyanines. Spectrochimica Acta Part A:

Molecular and Biomolecular Spectroscopy, 67(5):1232–1246, 2007.

[114] T. Takenobu, T. Takano, M. Shiraishi, Y. Murakami, M. Ata,

H. Kataura, Y. Achiba, and Y. Iwasa. Stable and controlled amphoteric

doping by encapsulation of organic molecules inside carbon nanotubes.

Nature Materials, 2(10):683–688, 2003.

[115] J. Zhu, B. S. Shim, M. Di Prima, and N. A. Kotov. Transparent

conductors from carbon nanotubes lbl-assembled with polymer dopant

with π−π electron transfer. Journal of the American Chemical Society,

133(19):7450–7460, 2011.

[116] M.S. Roy, P. Balraju, Y.S. Deol, S.K. Sharma, and G.D. Sharma.

Charge-transport and photocurrent generation in bulk hetero junc-

tion based on chloro-aluminum phthalocyanine (clalpc) and rose bengal

(rb). Journal of Materials Science, 43(16):5551–5563, 2008.

[117] A. V. Soldatova, J. Kim, X. Peng, A. Rosa, G. Ricciardi, M. E. Kenney,

and M. A. J. Rodgers. Effects of benzoannulation and α octabutoxy

substitution on the photophysical behavior of nickel phthalocyanines:

A combined experimental and dft/tddft study. Inorganic Chemistry,

46(6):2080–2093, 2007.

[118] Y. Zhang, W. Ruan, Z. Li, Y. Wu, and J. Zheng. Dft study on the

influence of< i > meso< /i >-phenyl substitution on the geomet-

ric, electronic structure and vibrational spectra of free base porphyrin.

Chemical physics, 315(1):201–213, 2005.

[119] B. Shan and K. Cho. First principles study of work functions of single

wall carbon nanotubes. Physical review letters, 94(23):236602, 2005.

[120] J. Kong, N. R. Franklin, C. Zhou, M. G. Chapline, S. Peng, K. Cho,

and H. Dai. Nanotube molecular wires as chemical sensors. Science,

287(5453):622–625, 2000.

178

[121] S. Maji and S. Sarkar. Homo based two electrons and one-

electron oxidation in planar and nonplanar methoxy-substituted nickel

tetraphenylporphyrins. Inorganica Chimica Acta, 363(12):2778–2785,

2010.

[122] N. Bendiab, L. Spina, A. Zahab, P. Poncharal, C. Marliere, J.L. Bantig-

nies, E. Anglaret, and J.L. Sauvajol. Combined in situ conductivity

and raman studies of rubidium doping of single-wall carbon nanotubes.

Physical Review B, 63(15):153407, 2001.

[123] A. Ilie, S. Egger, S. Friedrichs, D. Kang, and M. L. H. Green. Correlated

transport and high resolution transmission electron microscopy inves-

tigations on inorganic-filled single-walled carbon nanotubes showing

negative differential resistance. Applied Physics Letters, 91(25):253124–

253124, 2007.

[124] A. Weber-Bargioni, A. Schwartzberg, M. Cornaglia, A. Ismach, J. J.

Urban, Y. Pang, R. Gordon, J. Bokor, M. B. Salmeron, D. F. Ogletree,

et al. Hyperspectral nanoscale imaging on dielectric substrates with

coaxial optical antenna scan probes. Nano letters, 11(3):1201–1207,

2011.

[125] A. Hartschuh, E. J. Sanchez, X. S. Xie, and L. Novotny. High-resolution

near-field raman microscopy of single-walled carbon nanotubes. Phys-

ical Review Letters, (9):095503–1, 2003.

179